Effects of sentence-structure complexity on speech initiation time and disfluency

https://doi.org/10.1016/j.jfludis.2012.12.002Get rights and content

Abstract

There is general agreement that stuttering is caused by a variety of factors, and language formulation and speech motor control are two important factors that have been implicated in previous research, yet the exact nature of their effects is still not well understood. Our goal was to test the hypothesis that sentences of high structural complexity would incur greater processing costs than sentences of low structural complexity and these costs would be higher for adults who stutter than for adults who do not stutter. Fluent adults and adults who stutter participated in an experiment that required memorization of a sentence classified as low or high structural complexity followed by production of that sentence upon a visual cue. Both groups of speakers initiated most sentences significantly faster in the low structural complexity condition than in the high structural complexity condition. Adults who stutter were over-all slower in speech initiation than were fluent speakers, but there were no significant interactions between complexity and group. However, adults who stutter produced significantly more disfluencies in sentences of high structural complexity than in those of low complexity.

Educational objectives: After reading this article, the learner will be able to: (a) identify integral parts of all well-known models of adult sentence production; (b) summarize the way that sentence structure might negatively influence the speech production processes; (c) discuss whether sentence structure influences speech initiation time and disfluencies.

Highlights

► Sentence complexity affects speech initiation time and disfluency in fluent adults and adults who stutter. ► Slower speech initiation time is due to the increased processing cost placed on the speech production system. ► An increase in sentence complexity leads to a greater degree of disfluencies in adults who stutter relative to fluent adults.

Introduction

Stuttering affects approximately 1% of the world's population at any given time, and approximately 5% will experience it at some point in their lives (Bloodstein & Ratner, 2008). Having this communication disorder is a potentially devastating experience, affecting speech, language, social, academic, and occupational function in adults. After decades of research, there is general agreement that stuttering is caused by a variety of factors and their interaction; no single factor has been identified for all people who stutter (Kent, 2000, Smith and Kelly, 1997). Main factors appear to center around language formulation (e.g., lexical retrieval, syntactic construction, phonological encoding) and motor processes of speech, and there is suspicion that subgroups exist within the population of speakers who stutter (Kent, 2000, Tsiamtsiouris and Cairns, 2009, Watson et al., 1991). There are at least two reasons why it is important to increase our understanding of the different factors that lead to or maintain stuttering. First, from a theoretical perspective, we need to improve our understanding of the speech and language processes that are unique to people who stutter and the potential effects that the various components may have on fluency. Second, understanding these processes has the potential to influence the way we evaluate and treat stuttering in clinical settings. The current study investigated the language formulation basis of stuttering. In particular, the aim was to assess the effects of sentence structure on variables such as speech initiation time and disfluency. In the sections that follow, we describe what is known about sentence construction mechanisms, and, in particular, the syntactic skills of individuals who stutter.

The generation of sentence structure is an integral part of all well-known models of adult sentence production (Bock and Levelt, 1994, Garrett, 1988, Levelt, 1989). The first step of sentence construction is the formulation of an abstract idea. Then, sentence planning mechanisms must select the correct words (from the speaker's lexicon) and create the correct syntactic structure (based on the speaker's grammar) to convey the desired sentence meaning (Fernandez & Cairns, 2010, chap. 5). Grammatical encoding can be divided into functional processing and positional processing (Bock and Levelt, 1994, Garrett, 1988). Functional processing involves lexical and lemma selection. Information associated with a lexical item involves conceptual properties (domestic animal, four legs, fur, etc.), whereas lemmas contain grammatical information such as form class (content or function word, noun, verb, etc.) and subcategorical restrictions (whether a verb can be followed by an object, a clause, etc.). After the selection of appropriate lemmas, functional assignment calls for assigning syntactic relations or grammatical functions to the words (Bock & Levelt, 1994). In other words, the speaker's sentence production mechanism must specify the words that will serve as the subject of the sentence and which, if any, will serve as predicates of various kinds. Additional processing is required after the selection of lemmas, as the elements of an utterance need to be organized into a syntactic structure. This occurs in a subcomponent of positional processing, called constituent assembly. The result of these operations is a sentence structure of hierarchically ordered syntactic elements based on the speaker's grammar. This structure, together with the selected lexical items, conveys the meaning of the sentence. Despite the automaticity of these steps, sentence planning incurs processing costs, which vary with the complexity of the sentence. As the planned sentence is sent to the speech production system, complexity interacts with the motor systems involved in the actual production of the sentence (Kleinow and Smith, 2000, Namasivayam and Van Lieshout, 2011). Processing cost arises from the deployment of cognitive (psycholinguistic planning and execution) and physical (motor planning and execution) resources by the speaker (Kleinow and Smith, 2000, Tsiamtsiouris and Cairns, 2009).

Support for the view that speech and language factors are centrally implicated in stuttering comes from two types of studies. The first type involves the measurement of brain activity in specific areas in response to language processing. Such studies, in the absence of overt speech, demonstrate atypical semantic and syntactic processing in adults who stutter when compared to adults who do not stutter (Cuadrado and Weber-Fox, 2003, Weber-Fox, 2001, Weber-Fox et al., 2004). Other studies show atypical brain functions in areas related to speech, language, and auditory processing, supporting the hypothesis that language processing operates differently in individuals who stutter than in individuals who do not (Braun et al., 1997, Fox et al., 1996, Fox et al., 2000, Ingham et al., 2000, Preibisch et al., 2003). Another type of study experimentally manipulates linguistic variables and examines the effects of these variables (Kleinow and Smith, 2000, Melnick et al., 2003, Pellowski and Conture, 2005, Ratner and Sih, 1987, Silverman and Ratner, 1997) on characteristics of speech. A small number of these studies have focused on syntactic variables such as sentence structure in order to study the effects of increased processing cost on aspects of speech production such as speech initiation (Anderson and Conture, 2004, Tsiamtsiouris and Cairns, 2009).

One of the first studies (Johnson, 1966) to investigate the effects of syntactic complexity on speech initiation time involved sentences that differed according to the branching structure or hierarchical tree of the utterance. Some sentences had a more right branching structure or fewer syntactic nodes, and other sentences had more nodes and a more complex branching structure. This early study (Johnson, 1966) found that syntactic complexity affects speech initiation time, thereby supporting the notion that there are processing costs associated with the generation of sentence structure.

Tsiamtsiouris and Cairns (2009) employed Ferreira's (1991) experimental paradigm to measure the processing cost of sentence structure generation in adults who stutter and to test the hypothesis that this cost would be higher for adults who stutter than for adults who do not stutter. Participants saw a sentence, committed it to memory, and then verbally produced it after the sentence was removed and a cue prompted the person to start speaking. Sentence initiation time was measured; sentences were of either high or low syntactic complexity. The high complexity sentences contained a relative clause in the subject Noun Phrase (NP) and an additional relative clause after the main verb; the low complexity sentences were identical except for the fact that there was no relative clause in the subject NP. Instead, low complexity sentences had adjectives that preceded the subject. Length was controlled, as both sentence types were approximately 20 syllables long. Results for both groups of speakers revealed that speech initiation time was greater for high complexity sentences than for low complexity sentences. However, the difference was significantly greater for the people who stutter than for the fluent control group. These results confirmed the authors’ hypothesis that grammatical encoding costs would be more detrimental to adults who stutter than to the fluent participants, giving support to the theory that psycholinguistic factors are implicated in stuttering. Logan (2003) conducted a study with a similar experimental paradigm, but failed to show the differential processing cost of grammatical complexity for adults who do and do not stutter. The crucial difference between his sentences and those of Tsiamtsiouris and Cairns (2009) was that his were shorter and simpler. This suggests that there is a cost threshold that must be exceeded in order to observe a sentence structure complexity effect on speech initiation time in adult speakers; after this threshold is crossed, adults who stutter are more adversely affected than adults who do not stutter.

Silverman and Ratner (1997) conducted a study that had some methodological similarities to Logan (2003) and Tsiamtsiouris and Cairns (2009). These studies manipulated sentence structure complexity by incorporating a relative clause into the subject NP, defining this sentence type as high complexity. Sentences without relative clauses were considered lower in complexity. A major difference among the studies was the overall complexity level in terms of sentence structure and length. Tsiamtsiouris and Cairns’ (2009) sentences had two embedded relative clauses and approximately 20 syllables, whereas Silverman and Ratner's (1997) and Logan's (2003) had only one relative clause and approximately 11 syllables. All three studies controlled for sentence length. Another difference was the dependent variables. Instead of speech initiation time, Silverman and Ratner (1997) analyzed disfluencies and inaccuracies. Both speaker groups in Silverman and Ratner's study (1997) produced significantly more of these when imitating the high complexity sentences than when imitating the low complexity sentences. As in the Logan (2003) study, there was not a significant difference between speaker groups, reinforcing the hypothesis that sentences must exceed a threshold of complexity before there is a greater impact of complexity on adults who stutter than adults who do not stutter. However, the fact that both speaker groups produced more disfluencies and inaccuracies in the high relative to the low complexity sentences constitutes support for the hypothesis that increased complexity of the sentence structures creates an increased processing cost, leading to increased stress on the speech production system in all speakers.

Early work with typical adult speakers has connected normal disfluencies with the complexity of the speech planning task and the onset of clauses (Beattie, 1980, Boomer, 1965, Butterworth, 1980, Ford, 1978, Goldman-Eisler, 1972). This has led to the belief that the beginnings of clauses are critical planning junctures and areas of increased processing cost for children and adults (Bernstein, 1981, Bock, 1996, Garrett, 1982, McDaniel et al., 1998, McDaniel et al., 2010, McKee and McDaniel, 2001, Wall et al., 1981). In children, normal disfluencies at these points seem to be related to the ease with which fluent speakers can generate sentence structure (Wijnen, 1990, Wijnen, 1992, Rispoli and Hadley, 2001).

Research with people who stutter has shown us that sentence structure has the potential to negatively influence the speech production processes (Kleinow and Smith, 2000, Tsiamtsiouris and Cairns, 2009). There are a number of explanations for the increased sensitivity of people who stutter to sentence length and complexity (Starkweather and Gottwald, 1990, Wall et al., 1981). The beginnings of sentences and clause boundaries have shown an increased likelihood of stuttering (Brown, 1938, Brown, 1945, Silverman and Williams, 1967). Wall et al. (1981) showed a significant relationship between clause boundaries and location of stuttered events in children. Moreover, if sentence length is held constant, speakers who stutter have a tendency to exhibit more stuttering on the initial clause of a sentence than on the same clause in the final position of a sentence (Jayaram, 1984). A plausible explanation for these observations is that as speakers plan aspects of utterances and clauses prior to actual speech articulation (Garrett, 1975, Levelt et al., 1999), processing costs associated with this activity tax or overload the resources available for fluent speech production (Bosshardt, 1995, Starkweather and Gottwald, 1990, Yaruss, 1999). In other words, fluency breakdowns may occur if demands exceed a speaker's threshold or capacity for fluent speech. This is reminiscent of Rispoli's (2003) suggestion that children have a comfort zone in which it is easy to plan and produce linguistic information. If there are linguistic demands beyond the comfort zone, fluency can be compromised. In other words, if the cost to plan and produce syntactic structures is high, then construction of a sentence with complex structure has the potential to overwhelm the speech production system (Kleinow & Smith, 2000). The outcome is a breakdown in the coordinated functioning of the various speech and language systems that are essential for speech motor stability, initiation of speech, and fluency (Garrett, 1982, Kleinow and Smith, 2000, Namasivayam and Van Lieshout, 2011, Rispoli and Hadley, 2001, Tsiamtsiouris and Cairns, 2009).

Another line of inquiry has focused on whether people who stutter have an innate limitation (e.g., lower end of normal) of the speech motor control system and how this system is influenced by interactions with cognitive, linguistic, emotional, speech motor tasks (Peters, Hulstijn, & Van Lieshout, 2000). According to this model, an increase in processing cost or task complexity taxes a vulnerable speech motor control system and requires a speaker to use additional resources from a finite pool of central resources (Smits-Bandstra and De Nil, 2007, Van Lieshout et al., 2004). The strain that is placed on a vulnerable system can lead to speech motor control instability and an increase in disfluencies and other speech-related changes. It should be noted that in this model the speech motor control system of people who do not stutter is also taxed when there is an increase in processing cost; however, the main difference is that these individuals can handle the demands without a deleterious impact on fluency (Namasivayam & Van Lieshout, 2011).

In the current study, our goal was to create two lists of low and high complexity sentence types while controlling for length. It was hypothesized that sentential complexity would influence speech initiation time and disfluency in adults who do and do not stutter. A review of the psycholinguistic literature revealed an abundance of studies that investigated the impact of different sentence types on comprehension, but very few studies that examined the impact on production. Therefore, we will briefly discuss sentence types used in comprehension studies, as these sentence types will form the basis for our production study. Comprehension complexity can arise from many sources and is relevant to address particular theories of the operation of the structural processor (the parser) of the hearer. The parser creates a structural analysis of a heard sentence, which is necessary in order to recover the meaning encoded by the speaker (Fernandez & Cairns, 2010, chap. 7). Comprehension complexity can arise from increasing demands on sentence integration and working memory; processing elements that depart from canonical subject-verb-object order of grammatical relations; and processing elements with long distance dependencies. All of these factors increase the processing costs on the comprehension system. Below we describe different sentence types (low complexity paired with high complexity) that have been demonstrated to increase processing costs; these sentence types were used in the current production study.

A sentence that contains a clause (in this case a relative clause) is of greater complexity than one that does not. This is because the presence of a clause creates a constituent in the sentence structure not present in a sentence without a clause. Literature cited above implicates an association between clauses and disfluencies. Our first complexity pair contrasts a sentence with a relative clause in its subject Noun Phrase (e.g., “The plane that was leaving climbed higher into the sky that was clear and blue”) to one that does not (e.g., “The small but mighty plane climbed higher into the sky that was clear and blue). These sentence types have been used in two studies, Ferreira (1991) and Tsiamtsiouris and Cairns (2009), both measuring speech initiation times, the latter also measuring stutter-like disfluencies.

Sentences containing subject–object relative clauses (e.g., “The boy who the cat scratched cried”) are more costly to process than those containing object–subject relative clauses (e.g., “The boy petted the cat who purred.”). These sentence types have been studied extensively in comprehension, demonstrating the greater comprehension costs associated with subject–object relative clauses (Carpenter et al., 1994, Ford, 1983, Gibson, 1998, Holmes and O’Regan, 1981). In the more complex sentence type, containing subject–object relative clauses, the subject Noun Phrase (“the boy” in our example) is modified by a relative clause that intervenes between it and its verb (“cried”) and plays two grammatical roles, the subject of “cried” and the object of “scratched” (between which four words intervene). In the less complex sentence type, containing an object–subject relative clause, the relative clause modifies the object (“the cat”), which plays two grammatical roles, the object of “petted” and the subject of “purred;” both can be processed simultaneously, as there are no intervening words. Another major difference between the two sentences is that in the subject–object relative “the boy” is the object of “scratched,” giving the relative clause a non-canonical object–subject-verb order of grammatical relations, which is not the case with the other sentence. There are, then, a number of reasons why processing costs are greater for sentences containing a subject–object relative clause than those containing an object–subject relative clause.

It is more difficult to process a sentence that consists of an initial subordinate clause, followed by a main clause (e.g., “When the plane landed on the ground, the captain welcomed everyone to New York.”) than one that consists of an initial main clause followed by a subordinate clause (e.g., “The captain welcomed everyone to New York when the plane landed on the ground.”). There is abundant evidence from the comprehension literature that initial subordinate clauses are processed less completely than are initial main clauses (Bever and Townsend, 1979, Garnham et al., 1998, Jarvella, 1971, Townsend and Bever, 1978). This is because a main clause does not depend on a following subordinate clause for its meaning, therefore it can be completely processed. A subordinate clause depends on the main clause for its meaning, so it cannot be completely processed until the main clause is received. Anytime incompletely processed information must be held in memory processing costs are incurred. For this reason, sentences with a subordinate–main clause word order incur more processing costs than do sentences with main–subordinate order.

Sentences in the passive voice (e.g., “The small dog was chased by the black cat down the busy city street.”) incur more processing costs than do those in the active voice (e.g., “The black and white cat chased the small dog down the busy city street.”). Passive sentences have been studied throughout the history of psycholinguistics because they are a classic example of a sentence that violates the English canonical (standard) subject-verb-object order of grammatical elements. In this structure the object of the verb has moved to fill the subject position of the sentence, forming an argument chain between the object and subject positions. As in subject–object relative clauses, a single Noun Phrase serves simultaneously as subject and object of a verb. Passive sentences are known to be produced very late by English speaking children (Horgan, 1978); they are also difficult for young children to comprehend (Strohner & Nelson, 1974). Studies also show that passive sentences are more costly for adults to process than are active sentences (Gough, 1966, Mehler, 1963).

In both verb phrase complements (e.g., “The teacher told the students in the room to sit down and be quiet.”) and adverbial adjunct clauses (e.g., “The child played with the toy while waiting for his mom to return.”), clausal subjects (of “to sit” and “waiting”, respectively) must be inferred from the structure of the sentence. There has been extensive study of the comprehension of these forms by children (Cairns et al., 1994, Hsu et al., 1985); VP complements are correctly interpreted at a much earlier age than are adverbial adjuncts. In fact, young children incorrectly interpret the object of the adverbial adjuncts as being the object of the main clause, in an analogy to the VP complements. There is no evidence that adults incur greater processing costs with adverbial adjuncts than with VP complements; however, the difference in children's interpretation of the two structures suggested to us that those with adverbial adjunct clauses may be of greater complexity, so we included them in our study.

In summary, research with people who stutter has shown us that sentence structure has the potential to negatively influence the speech production processes, and a number of studies have identified a variety of low and high complexity sentence types that influence processing cost. Most of these studies that manipulated complexity to examine processing cost have focused on comprehension, and it remains to be seen whether these sentence types show a higher processing cost during sentence production. The hypothesis that high complexity sentences will create longer speech initiation times and more disfluencies than low complexity sentences has important implications for psycholinguistic theory. Confirmation would suggest a processing connection between the comprehension and production systems. In addition, if effects of complexity are more pronounced for adults who stutter than for people who do not stutter, this result will constitute additional evidence for theories of stuttering that implicate sentential complexity in people who stutter.

The research questions for this study were the following: (1) Do all speakers initiate low complexity sentences significantly faster than high complexity sentences? (2) Do adults who stutter initiate all sentences significantly more slowly than adults who do not stutter, regardless of complexity? (3) Does sentence complexity have a greater impact on the speech initiation time of adults who stutter than on adults who do not stutter? (4) Do all speakers produce low complexity sentences with significantly fewer disfluencies than high complexity sentences? (5) Does sentence complexity have a greater impact on speech disfluencies in adults who stutter than adults who do not stutter?

Section snippets

Participants

Participants were 21 adults who stutter (M = 29.33 years; SD = 6.85 years) and 21 adults who did not stutter (M = 29.42 years; SD = 7.40 years). Every individual in the experimental group (17 males, 4 females) was matched for age (±3 years), gender, and education with an adult who did not stutter. All individuals had at least some college education.

Both groups of speakers were required to have American English as their primary language and general language skills that passed several screening measures.

Speech initiation time analyses

Responses that were produced without disfluencies and errors were analyzed for speech initiation time. Discarded trials resulted in the loss of 5.7% of trials for the fluent adults and 14.7% for adults who stutter. This difference was not a surprising outcome, as adults who stutter had a higher number of sentences with disfluencies. Table 1 displays mean speech initiation times and standard deviations by condition and group. A mixed ANOVA revealed significant main effects of group F(1,40) = 

Discussion

We hypothesized that the production of structurally complex sentences would incur a high processing cost and that this high cost would have an impact on speech characteristics such as initiation time or disfluency. We begin our discussion by reviewing the effects of complexity on speech initiation time and disfluency, and then turn to our thoughts about the theoretical implications of its impact on adults who stutter.

Conclusions

The present investigation focused on whether sentences of high structural complexity relative to low structural complexity would produce greater production processing costs for fluent adults and adults who stutter, and whether this difference would be exacerbated in the latter group. Results showed that both groups of speakers initiated sentences significantly faster in the low structural complexity condition than in the high structural complexity condition, but adults who stutter were over-all

Jim Tsiamtsiouris, Ph.D., CCC-SLP is an Assistant Professor in the Department of Communication Disorders at William Paterson University. His primary research focus is the contribution of psycholinguistic variables on speech fluency.

References (79)

  • K. Logan

    The effect of syntactic structure upon speech initiation times of stuttering and nonstuttering speakers

    Journal of Fluency Disorders

    (2003)
  • J. Mehler

    Some effects of grammatical transformation on the recall of English sentences

    Journal of Verbal Learning and Verbal Behavior

    (1963)
  • C. Preibisch et al.

    Evidence for compensation for stuttering by the right frontal operculum

    NeuroImage

    (2003)
  • M. Smith et al.

    Syntactic priming in spoken sentence production—An online study

    Cognition

    (2001)
  • S. Smits-Bandstra et al.

    Sequence skill learning in persons who stutter: Implications for cortico-striato-thalamo-cortical dysfunction

    Journal of Fluency Disorders

    (2007)
  • C.W. Starkweather et al.

    The demands and capacities model II: Clinical applications

    Journal of Fluency Disorders

    (1990)
  • D. Townsend et al.

    Interclause relations and clausal processing

    Journal of Verbal Learning and Verbal Behavior

    (1978)
  • M. Wall et al.

    Syntactic influences on stuttering in young child stutterers

    Journal of Fluency Disorders

    (1981)
  • B. Watson et al.

    Linguistic performance deficits in stutterers: Relation to laryngeal reaction time profiles

    Journal of Fluency Disorders

    (1991)
  • F. Wijnen

    Incidental word and sound errors in young speakers

    Journal of Memory and Language

    (1992)
  • J. Anderson et al.

    Sentence-structure priming in young children who do and do not stutter

    Journal of Speech, Language, and Hearing Research

    (2004)
  • G. Beattie

    The role of language production processes in the organization of behavior in face to face interaction

  • T. Bever et al.

    Perceptual mechanisms and formal properties of main and subordinate clauses

  • O. Bloodstein et al.

    A Handbook on Stuttering

    (2008)
  • K. Bock

    Language production: Methods and methodologies

    Psychonomic Bulletin & Review

    (1996)
  • K. Bock et al.

    Language production: Grammatical encoding

  • D. Boomer

    Hesitation and grammatical encoding

    Language and Speech

    (1965)
  • Bosshardt, H. (1995). Syntactic complexity, short-term memory, and stuttering. Paper presented at the Annual Convention...
  • A.R. Braun et al.

    Altered patterns of cerebral activity during speech and language production in developmental stuttering: An H2 150 positron emission tomography study

    Brain

    (1997)
  • S.F. Brown

    Stuttering with relation to word accent and word position

    Journal of Abnormal Social Psychology

    (1938)
  • S.F. Brown

    The loci of stuttering in the speech sequence

    Journal of Speech Disorders

    (1945)
  • B. Butterworth

    Some constraints on models of language production

  • H. Cairns et al.

    A longitudinal study of principles of control and pronominal reference in child English

    Language

    (1994)
  • P. Carpenter et al.

    Working memory constraints in comprehension: Evidence from individual differences, aphasia, and aging

  • E.G. Conture

    Stuttering: Its nature, diagnosis, and treatment

    (2001)
  • E.G. Conture et al.

    Communication-emotional model of stuttering

  • E. Cuadrado et al.

    Atypical syntactic processing in individuals who stutter: Evidence from event-related brain potentials and behavioral measures

    Journal of Speech, Language, and Hearing Research

    (2003)
  • E.M. Fernandez et al.

    Fundamentals of psycholinguistics

    (2010)
  • Ford, M. (1978). Planning units and syntax in sentence production. Doctoral dissertation. University of Melbourne,...
  • Cited by (0)

    Jim Tsiamtsiouris, Ph.D., CCC-SLP is an Assistant Professor in the Department of Communication Disorders at William Paterson University. His primary research focus is the contribution of psycholinguistic variables on speech fluency.

    View full text