The position taken in this article is that late developing sounds are not only relevant to child phonological development, but may also indicate either the motoric difficulty or the pre-motor planning of different sounds for speakers of all ages. LEC and CS may be acquired late if the articulators require time to mature (see Schwartz, 1988, for a discussion on motor limitation and language development). However, these low-level (physical or biological) limitations are hard to be distinguished from high-level factors (planning or representation) in phonological development (c.f. Smith, 1988). The comment on historical sound change by Slobin (1977) also reflects the view that “forms whichare late to be acquired by children are presumably also relatively difficult for adults to process” (p. 194; cited in Locke, 1983). Adults have been reported to avoid words with consonant sequences that they find difficult (Locke, 1983; Schwartz, 1988). The current study tests if LEC and CS are particularly difficult for people who stutter and how this varies between different age ranges. Sander's (1972) study based on data from children up to eight years old and other studies on avoidance of words were conducted on children still undergoing phonological development. The current study uses age ranges up to 40. It is predicted that sounds which are difficult for young children will continue to be problematic for adults who stutter.

Both LEC and CS involve consonants and consonants are known to be related to high incidence of stuttering for people who stutter (Brown, 1945). Consequently, Throneburg et al. (1994) investigated whether words containing CS and LEC have high levels of motor difficulty that results in these sounds being stuttered in a group of children. Howell and Au-Yeung (1995a) investigated these factors in children who stutter over three age groups (2 to 6, 6 plus to 9 and 9 plus to 12). No dependence of frequency of stuttering on phonological difficulty was found by either Throneburg et al. (1994) or Howell and Au-Yeung (1995a).

In both these studies, each word was assessed in terms of whether it contained one or more of the three factors that indicate phonological difficulty. A limitation in these studies is that no distinction was made concerning the position where an LEC or CS occurred within a word. Consequently even for monosyllable words, CS and LEC factors can be distributed over different word positions in several ways: CSs can be word initial (e.g., ‘quick’), or non-initial (e.g., ‘found’); LEC can also be word initial (e.g., ‘like’) or non-initial (e.g., ‘fail’); A word containing CS and LEC can have both CS and LEC word initial (e.g., ‘tried’, or ‘school’), both CS and LEC can be non-initial (e.g., ‘helps’) or CS and LEC can be at different positions (e.g., ‘quiz’ has CS initial and LEC non-initial and ‘lent’ has these factors in the reverse positions).

Two observations emphasize the importance of performing the analyses with regard to the position that the phonological factors occupy in a word. First, stuttering almost always occurs on the first phones of a word (Brown, 1945; Wingate, 1982, 1988). Second, there may be some interaction between CS and LEC such that when they occur in the same word position, this enhances their difficulty. Thus, a word like “school” that has a CS and LEC that occur in word-initial position would be expected to cause more difficulty than “quiz” where the CS occurs before the vocalic segment and the LEC after it. Taken together, these two observations suggest that if CS and LEC both occur before the first vowel in a word, the word would be highly prone to being stuttered (what is referred to as focal difficulty). The experiment reported below set out to determine whether incidence of stuttering depends on CS and LEC positions and, in particular, whether words with focal difficulty have a higher chance of being stuttered than other CS/LEC combinations within a word.

Examination of these factors over age groups is also important because differential influences would be expected according to the speaker's age. There are two alternative views about how such influences vary with speaker's age. First, when speakers make their initial attempts at sounds with LEC and CS, they may have difficulty. Once they have acquired the ability to produce these sounds, they are no more difficult than others that do not involve these factors. In this case, CS and LEC would lead to production problems in early development but not at later ages.

The second view is that sounds that are difficult early in development continue to be motorically difficult for older speakers, particularly those who persist in their stuttering. Van Lieshout, et al. (1995) offer some support for this view. They investigated EMG activity in fluent adult speakers on words that contained linguistic factors known to affect frequency of stuttering. They found that words that are known to affect frequency of stuttering in adults resulted in increased EMG activity in their speakers. Our interpretation of these results is that sounds with a high likelihood of being stuttered are problematic even for fluent speakers into adulthood. This view is similar to one that Throneburg et al. (1994) discuss about how CS and LEC could give rise to motor difficulty. In this, they suggest that CS and LEC may reflect the difficulty-loading on the supposedly inferior peripheral articulatory system of children who stutter. The alternative viewpoint they raise is that these factors could cause difficulty in producing sounds containing them due to high level of demand on pre-motor planning. Though Van Lieshout et al. (1995) did not investigate CS and LEC, if sounds with these properties are acquired late because they are motorically difficult for speakers in general, it would be expected that they would continue to cause problems into adulthood as with words containing the factors van Lieshout et al. (1995) investigated.

The more detailed breakdown of both CS and LEC by word position showed that words containing neither of these factors declined in usage over age groups. Once again, this suggests that simple phonological structures are used less frequently as speakers get older. The decrease in usage of this category over age groups is dramatic. The reduction due to the decline in content words with neither an LEC nor a CS is offset by increases in words with more complex phonological structures. It seems likely that this reflects changes in vocabulary which would be expected to occur over the entire lifespan.

Previous studies carried out by Throneburg et al. (1994) and by Howell and Au-Yeung (1995a) found no effect of either LEC or CS on the occurrence of stuttering. The current study, however, showed that the phonological factors of LEC and CS affected frequency of stuttering when word positions of these factors were considered in Analysis 1. Speakers from all age groups stuttered significantly more on content words that starting with CSs than those that contained no CS at all. Moreover, the youngest speakers (group 1) stuttered significantly more on words starting with LECs. On the other hand, Analysis 2 showed that no effect of the two phonological factors, LEC and CS, was found on the stuttering of function words. This is compatible with the suggestion put forward by Au-Yeung and Howell (1998) that word external properties trigger stuttering on function words. At the same time, word internal factors, in this case LEC and CS, affected stuttering in all speakers. This suggests that content words and function words need to be treated differently when studying their linguistic effect on stuttering.

When the interaction of CS and LEC was considered in Analysis 3, variation in stuttering ratio was investigated over the five LEC/CS classes where there were sufficient occurrences (categories 1, 2, 3, 5, and 9). Significant differences over word categories occurred for speakers in age group 2 and group 3. When content words contained neither an LEC nor a CS (category 1), the words were stuttered less and can be regarded as phonologically easy. Similarly category 3 can be considered easy since many of this class of content words include regularly used plural word endings like “-s.” The category of words with focal difficulty, on the other hand, showed an increased chance of being stuttered.

In normal phonological development, LECs are used correctly much earlier than CSs by the average child. Sander (1972) found that the age of customary articulation for LECs was around 36 months. His study, however, did not distinguish word position for LECs. Prather, Hedrick, and Kern (1975) studied the acquisition of the consonantal sounds at both word initial and word final position and Sander's LECs at word initial position were still found to be mastered by children at around 36 months. At word final position, a slightly different pattern appears. On the other hand, Wellman et al. (1931) reported that normally developing children could not produce CSs correctly before 60 months except the cluster /kw/. Smit (1993) also noted that adults learning a second language also found CSs difficult if their first language did not allow the strings. Thus, it is reasonable to assume that CSs are more taxing for the articulators than LECs for both fluent and non-fluent speakers.

In the current study on speakers who stutter, the results for the youngest group of speakers suggest that, before age 12, children who stutter found words starting with CS or LEC difficult. These two factors seem to operate independently in this age group. The focally difficult class of words (CSi-LECi) did not come out as difficult for these speakers in Analysis 3. An inspection of Figure 9 showed that there is an increasing trend for a higher proportion of this class of words to be stuttered as age increases. Speakers in group 1 found content words with LEC at initial position (CSx-LECi and CSi-LECi, where the latter is also focally difficult) more difficult than other content words although not significantly so.

For speakers in group 2, among words without CS at all (CSx), words whose onset involved LECs (CSx-LECi) were more dysfluent than words without LEC at the onset (CSx-LECx and CSx-LECn). The same result no longer existed for speakers in group 3. An inspection of Figure 9 showed that the stuttering ratio of the CSx-LECi is decreasing gradually with age. A significant difference between the CSx-LECi word group and the CSi-LECi word group emerged for the first time in the adult speakers. As age progressed, the speakers might have a better mastery of words starting with LECs and stuttered less on these words. At the same time, the CSn-LECn words, in which CS and LEC may co-occur at non-word-initial positions, were now more dysfluent than CSx-LECx words.

The more difficult words involving CSs at word initial position, however, remains difficult for people who stutter even when they get older. This shows that there may still be a deficit in the articulators in the older speakers whose stuttering persisted or that the therapies they received were not directed to deal with the articulation of consonant strings. Alternatively, the phonetic planning of such words in preparation for motor execution might have remained difficult for adult speakers who stutter.

It was not possible to fully test for ‘focally’ in this study. The main drawback is that only the CSi-LECi class has words with the CSi element. There were not enough occurrences of words containing CSi but not LECi (CSi-LECn and CSi-LECx) in the speech data because most CSi involve LEC in English.

The results of this study show that the two phonological factors, LEC and CS, have an impact on the likelihood that a word would be stuttered or not. The effect depends on the age of the speakers who stutter. It has also been shown that such influence would only apply to content words and at word initial positions but not to function words at any word position. The theory developed by Au-Yeung, Howell, & Pilgrim (1998) and Howell, Au-Yeung, & Sackin (1999) can be used to explain our current findings. The theory explains why the linguistic factors that make a word prone to being stuttered change across ages. The theory, as originally formulated, attempted to account for why stuttering in children occurs predominantly on function words whereas stuttering in older speakers occurs mainly on content words. Stuttering at an early age has similarities with the dysfluencies fluent speakers of all ages produce: Maclay and Osgood's (1959) data on fluent adults show that function words are repeated frequently. Clark and Clark (1977) accounted for this type of dysfluency by proposing that the subsequent word is not prepared for articulation and so the speaker hesitates or repeats the function word or words that have been prepared until the phonetic plan for the word that is not ready has been completed. Au-Yeung, Howell, & Pilgrim (1998) pointed out that the words for which the phonetic plan is not available will tend to be a phonologically more complex content word. Thus, in an example like “and, and, Christine,” the plan for the content word “Christine” is not available and so the speaker delays by repeating the function word “and.” Though this has been described as a mechanism that fluent speakers adopt, it is consistent with the observation that children who stutter exhibit a high proportion of stuttering on clauses starting with the function word “and” (Howell & Au-Yeung, 1995b). Howell, Au-Yeung, & Sackin (1999) provide evidence that they interpret as showing the increased incidence of stuttering on content words in adults is due to these speakers overriding the delaying on the function word. The speakers then have to attempt the content word that is not fully prepared for articulation. This results in the speakers experiencing difficulty in producing content words. Fluent speakers continue to repeat function words when the plan is not available and, consequently, do not encounter this problem. They also propose that the difficulties experienced on content words by people who stutter becomes an automatized response and this makes the problem resistant to removal without treatment (Shiffrin & Schneider, 1977).

Use of the delaying mechanism on function words in children who stutter has two effects: a) The repeated function words are less likely to contain CS and LEC (as shown in Analysis 2); b) When they delay on the function word, they are able to avoid the content words that are more likely to contain phonologically more difficult components such as CS and LEC. Since adults who stutter, on the other hand, attempt content words that are not fully prepared for articulation, the automatization of stuttering on such words is likely to increase the influence of phonological factors like CS and LEC. This theory predicts that as speakers drop repetition of function words, they are obliged to attempt partly prepared content words. This could lead to an increased frequency of stuttering in two ways: First, focally difficult content words may be the ones that are more inclined to have errors in their plan (Levelt, 1989). Second, attempting to execute partly prepared sounds leads to dysfluencies. At present, there are no data available that distinguishes between these alternatives.

In other future work, it will be necessary to look at the exact composition of the CS and its relationship to stuttering. To explain, inspection of the data shows that syllable-initial and syllable-final CSs are usually different from one another. For example, the consonant strings ‘nt’, ‘nd’, ‘lt’, ‘ŋz’, etc. found at syllable-final position are not allowed at syllable-initial position. The same is true for other consonant strings that can only be syllable-initial in English, for example, ‘fr’, ‘str’, ‘kw’, etc. Other strings occur at both initial and final position of a syllable, for example, ‘st’, ‘sk’, etc. This asymmetry alone could account for the word-initial effect of CS. Thus, it will be necessary to establish whether word-initial position of the CS per se, or the inherent difficulty of the CSs that occur in initial position induce stuttering by looking at individual combinations of CSs. Another relevant observation that emphasizes the need for looking at individual CSs is that the phonotactic rules of English lead to most syllable-initial CSs including LECs, whereas LECs do not occur so frequently for syllable-final CSs. In order to carry out such a study, a lot of data is needed so that there are adequate occurrences of each CS type for each speaker.

In the present article, linguistic characterizations of phonological difficulty have been used. It is entirely possible that these abstract characteristics may be related to motor indices. There is evidence in the literatures that some particular sounds or sequences of sounds are more difficult for people who stutter than those who do not (Van Lieshout, et al., 1997). Further research should be conducted to test if words with CS and LEC factors occurring focally at word-initial position are particular motorically taxing for people who stutter.