Despite the expression "fluency" is much used in stuttering-related scientific literature, there is not an agreement about its definition [1, 3, 12]. The following observations suggest that fluency is not confined to the non-occurrence of stuttered disfluencies (part-word repetitions, initial prolongations, and blocks):

• Some people who stutter (PWS) may be able to avoid the explicit emergence of stuttered disfluencies through the voluntary and conscious use of common disfluencies: after the detection of a possible occurrence of a disruption in a word that would be uttered, the speaker uses a common disfluency until he or she foresees the word will be uttered without disruption. Although this strategy may work with a high level of success, people who use it may not always be considered fluent speakers.

• The so called "covert stuttering" is characterized by the frequent use of reformulations in the form (and, inevitably, in the content) of speech. Reformulations include word substitutions, changing word order in the phrase, and reformulations of whole sentences. Although there may be few stuttered disfluencies in the speech of people with covert stuttering, some of these people will not be considered fluent speakers.

• Some PWS may be able to avoid the explicit emergence of stuttered disfluencies by decreasing speech rate and increasing the frequency and duration of silent pauses. Although these are very useful strategies for reducing stuttering, if the resultant speech is too slow, with too too long or too many pauses, the speaker will hardly be considered as a fluent speaker.

• Some patients that apply so-called "speech techniques" can only use them with great effort and concentration. Although the speech of these patients usually presents few stuttered disfluencies, they may not be considered as fluent speakers.

The use of the expression "fluency" is not restricted to referring to the classical fluency disorders (stuttering and cluttering), but it is also appropriate to other language disorders, such as:

• People with specific language impairment are not usually considered as fluent speakers, although they do not usually present excessive stuttered disfluencies.

• The same is true for people with non-fluent aphasias (Broca aphasia, motor transcortical aphasia, and global aphasia). Stuttered disfluencies are not a problem for these patients, but many of them are certainly not considered as fluent speakers.

Fluency as a skill

What does fluency have in common with walking, driving, concentrating, and doing calculus? These are all examples of skills. Fluency is also a kind of skill, but one of a very specific domain -- language [8, 9, 13, 14]. Skills have two typical characteristics [16]:

• They are acquired little by little. It is not possible to acquire a skill instantaneously. Practice is necessary. And practice requires time. In the specific case of fluency, this means that to be fluent one has to talk fluently many times. So we understand why fluency tends to improve with time as the person practices -- it is because the skill is gradually strengthened.

• Skills tend not to be flexible, that is, to learn a specific skill does not usually make learning another easier. In the specific case of fluency, this means that being fluent in a face-to-face conversation does not necessarily imply being fluent in telephone conversations or in public presentations. Therefore, it is easy to understand why some people complain of fluency difficulties in some situations but not in others. Despite fluency having similarities in several speech situations, there are also differences that need to be identified if the person wishes to have more fluency in a specific situation.

Possible components of verbal fluency

As fluency has several components, evaluating fluency requires considering several aspects of speech.

1) Low frequency of disfluencies

Disfluencies used to be seen as speech errors or defects, but not anymore, because one now knows that disfluencies give the speaker time to solve transitory difficulties related to "what" to say or "how" to say something [6, 15]. Disfluencies are part of the speech of all speakers (with and without communication disorders). The same psycholinguistic processes that generate fluency, generate disfluency. Perfect fluency is an abstraction, and only exists in professional or rehearsed readings and in spoken texts rehearsed or learned by heart. This is to say that disfluencies only disappear during the production of familiar, fixed, and automatic utterances [13].

There are two kinds of disfluencies [17]:

• Common disfluencies are present in the speech of all speakers. They are hesitation pauses, filled pauses (like "uh" and "um"), final prolongations, word repetitions, and false starts. In normally fluent speakers, there are at most 10% speech disfluencies [19].

• Stuttered disfluencies are typically present in the speech of PWS, but may be eventually observed in the speech of people who do not stutter. They are part-word repetitions, initial prolongations, and blocks. In normally fluent speakers, there are at most 3% of stuttered disfluencies in speech [19].

2) Low frequency of reformulations

Reformulations signal fragments considered inadequate by the speaker or by the listener that are repaired. Sometimes it is possible to observe the presence of reformulation markers, such as: "nay", "for that matter", or "that is". Reformulations include paraphrases (the corrected fragment is similar in content to the original one) and corrections (the corrected fragment is not similar in content to the original one) [6].

Normally fluent speakers show a low frequency of reformulations in speech, but, as far as we know, there are not numeric parameters for normality [6, 15].

3) Fluent silent pauses

Not all pauses in speech are the same. There are two types of silent pauses: fluent and hesitant [2]. Fluent pauses occur in strong syntactic boundaries (between clauses or phrases); there is also a significant difference between vocal frequency before and after the pause. Hesitation pauses occur in weak syntactic boundaries (inside phrases) and there is not a significant difference between vocal frequency before and after the pause.

There are basically three variables for analyzing pauses: frequency, duration, and use [1, 11]. Frequency is more important for the perception of fluency than duration, that is, the high frequency of pauses is more related to the perception of lack of fluency than the long duration of pauses. Besides, a native-like use of pauses is necessary for the perception of fluency.

Contrary to common sense, silent pauses do not happen onlyfor breathing. They are useful to physically disconnect clauses and phrases, which makes meaning establishment easier. However, when the speaker pauses, he/she can also take advantage of the opportunity to breathe [4].

In the speech of normally fluent speakers, about 70% of pauses are fluent and about 30% are hesitant. In addition, fluent pauses are, on average, 800 milliseconds longer than hesitation pauses [18].

It is also known that speakers who do pause at strong syntactic boundaries show fewer occurrences of disfluencies in speech. Thus, pauses are important instances for language planning and can be used to prevent difficulties in the production of utterances [18].

Normally fluent speakers show adequate pauses regarding frequency, duration, and use.

4) Speech rate

Speech rate refers to whether a speaker talks slowly, on average, or fast. It is usually measured in syllables per minute (a measure of articulatory programming) or in words per minute (a measure of information production) [3].

It is important to note that speech rate values change according to the sociolinguistic community. For that reason, there is not a universal value that could be applied to all speakers of all languages [7]. Thus, a certain value of speech rate considered average in a specific community can be taken as slow or fast in another. For example, Brazilian adults who live in the city of Sao Paulo and speak Portuguese, usually have a speech rate from 120 to 140 words per minute and from 220 to 260 syllables per minute [20].

Normally fluent speakers present comfortable speech rates, which are not too slow or too fast for their sociolinguistic community [5, 7].

5) Smoothness of speaking

Smoothness or ease of speaking refers to the lack of effort employed during speech production. Effort is usually evaluated from a qualitative and global point of view [1].

Effort is physically related to the pressure below the larynx, larynx tension, and tongue pressure [5]. Effort during speech production can also relate to semantic or cognitive difficulties (for instance, difficulties remembering).

Normally fluent speakers present low effort during speech production.

6) Grammatical skills

The ease a speaker has in applying word-building and word-combination rules during spontaneous speech is also important for the notion of fluency. Research about the relationship between grammar and fluency are only beginning, but there are clues that grammatical skills are important for fluency [3, 11, 13].

Adults without communication disorders do not usually face great difficulties in using their native language grammar. Nevertheless, fragments without articles, prepositions, or conjunctions (instances of agrammatism) are considered less fluent, because they show disruptions in the grammatical construction of utterances.

Normally fluent speakers present good grammatical skills and few instances of agrammatism in spontaneous speech.

7) Semantic complexity

Research about the relationship between semantics and fluency are also only beginning, but there are clues that semantic complexity affects fluency [3, 10, 13].

Semantic complexity depends on lexical and supralexical factors [3]. Regarding lexical units, factors that suggest greater complexity include the use of non frequent words and the restricted use of expressions with low semantic weight (for example: "thing", "stuff", "kind of", "you know"). As to supralexical units, factors that suggest greater complexity include low redundancy, structural cohesion, number and complexity of concepts, and semantic relations among concepts.

Normally fluent subjects talk with average or high semantic complexity.

All the seven factors discussed here converge to determine if speech is more or less fluent. Nevertheless, these factors also connect among themselves, creating a complex network. For example, disfluencies are more frequent if there are fewer fluent pauses, faster speech rates, and higher semantic complexity.

Although it is known that fluency results from many factors, we still need to agree about all these factors [1], to better specify each of them [3], to know which ones are central and which are peripheral to the notion of fluency [8], and to know how they relate to each other [3].

If it is clear that fluency includes much more than stuttered disfluencies. Fluency evaluation and the goals of speech therapy for stuttering should not be restricted to eliminating stuttered disfluencies. That is why I emphasize the relevance of this discussion about the concept of fluency, for if we know what our object is about, it will be easier to know where to go.

1. CHAMBERS, Francine. (1997). What do we mean by fluency? System 25 (4), 535-544.
2. CRUTTENDEN, Alan. (1994). The forms of intonation. In: Intonation. 4th ed. Cambridge: Cambridge University Press. p. 35-74.
3. FEYEREISEN, Pierre; PILLON, Agnesa & DE PARTZ, Marie-Pierre. (1991). On the measures of fluency in the assessment of spontaneous speech production by aphasic subjects. Aphasiology 5 (1), p. 1-21.
4. GROSJEAN, François; GROSJEAN, Lysiane & LANE, Harlan. (1979). The patterns of silence: performance structures in sentence production. Cognitive Psychology 11, p. 58-81.
5. INGHAM, Roger J. & CORDES, Anne K. (1997). Self-measurement and evaluating stuttering treatment efficacy. In: Curlee, Richard F. & Siegel, Gerald M. (eds). Nature and Treatment of Stuttering: New Directions. 2nd ed. Boston: Allyn and Bacon. p. 413-437.
6. KOCH, Ingedore G. Villaça & SOUZA E SILVA, Maria Cecília Párez de. (1996). Atividades de composição do texto falado: a elocução formal. In: Castilho, Ataliba Teixeira de & Basilio, Margarida (orgs). Gramática do Português Falado. Volume IV: Estudos Descritivos. Campinas: Editora da UNICAMP e São Paulo: FAPESP. p. 379-410.
7. LAVER, John. (1995). The temporal organization of speech: continuity and rate. In: Principles of Phonetics. Cambridge: Cambridge University Press. p. 534-546.
8. LENNON, P. (1990). Investigating fluency in EFL: a quantitative approach. Language Learning 40, 387-417.
9. LEVELT, Willem J. M. (1989). Speaking: from Intention to Articulation. Cambridge and London: The MIT Press and A Bradford Book.
10. MANSUR, Letícia Lessa & RADANOVIC, Márcia. (2004). Neurolingüística: Princípios para a Prática Clínica. São Paulo: Ediçöes Inteligentes.
11. RICHARDS, Jack; PLATT, John & WEBER, Heidi. (1985). Longman Dictionary of Applied Linguistics. Essex (England): Longman Group Limited.
12. RIGGENBACH, Heidi. (1991). Toward an understanding of fluency: a microanalysis of nonnative speaker conversations. Discourse Processes 14, p. 423-441.
13. SCARPA, Ester Miriam. (1995). Sobre o sujeito fluente. Cadernos de Estudos Lingüísticos 29, 163-184.
14. SCHMIDT, R. (1992). Psychological mechanisms underlying second language fluency. SSLA 14, 357-385.
15. SOUZA E SILVA, Maria Cecília Párez de & KOCH, Ingedore G. Villaça. (2002). Estratágias de desaceleração do texto falado. In: Kato, Mary A. (org). Gramática do Português Falado. Volume V: Convergências. 2nd ed. Campinas: Editora da UNICAMP. p. 329-340.
16. SQUIRE, Larry R. & KNOWLTON, Barbara J. (1997). Memory, hippocampus and brain systems. In: Gazzaniga, Michael S. (ed). The Cognitive Neurosciences. 4th ed. Cambridge: A Bradford Book. p. 825-837.
17. WINGATE, Marcel E. (1987). Fluency and disfluency; illusion and identification. Journal of Fluency Disorders 12, p. 79-101.
18. WINKWORTH, Alison L.; DAVIS, Pamela J.; ADAMS, Roger D. & ELLIS, Elizabeth. (1995). Breathing patterns during spontaneous speech. Journal of Speech and Hearing Research 38, p. 124-144.
19. YAIRI, Ehud. (1997). Disfluency characteristics of childhood stuttering. In: Curlee, Richard F. & Siegel, Gerald M. (eds). Nature and Treatment of Stuttering: New Directions. 2nd ed. Boston: Allyn and Bacon. p. 49-78.
20. ZACKIEWICZ, Daniela Verônica & ANDRADE, Claudia Regina Furquim de. (2000). Seis parâmetros de fluência. Revista da Sociedade Brasileira de Fonoaudiologia 5 (7), p. 59-64.

You can post Questions/comments about the above paper to the author before October 22, 2008.