METODOLOGÍA

This review has so far suggested that speech output skills, as measured by speech rate, is in some way implicated in the process o f verbally recalling the items to be remembered in STM tasks. If speech rate is a critical

measure o f STM performance, we should consider what aspect(s) o f speech output processing are being tapped by speech rate measures. It may not be the motoric skill o f rapid oral movement (motor execution) that allows speech rate to play an important role in predicting memory span. Rather, it may be the chain o f output processing from access to motor programs, through motor planning, to motor execution that is implicated.

As reported previously in this chapter, Henry (1994) found that speech rate correlated with memory span when the measure o f speech rate was based on the repetition o f pairs or triads of words in children aged 5, 7 and 11 years. However, when the speech rate measure was for the repetition o f a single word, this correlation was not significant. Henry (1994) suggests that the memory load o f retaining two or three words in STM affects speech rate. She also considers that a speech output planning facility may be a factor in both speech rate and STM span. This facility would be taxed more fully when repeating a sequence o f two or three words rather than a single word. If motor planning is influential in span performance, rather than motor execution, then repetition o f pairs or triads is the measure that would be more closely linked with span, than repetition o f a single word. Hulme et al. (1984) also found that speech rate for single words was more rapid than for groups o f three words, and pinpoints greater demands in programming and executing motor commands for sequencing o f three words rather than one.

Raine, Hulme, Chadderton and Bailey (1991) acknowledge that speech rate might be “an index o f some other more central verbal process, which in turn determines the efficiency o f STM performance” (p. 416). Raine et al.’s (1991) study investigated children, aged between 4 and 15 years, with a variety o f speech disorders. As speech difficulty might mitigate against

successful performance in any task requiring verbal output, the stimuli for recall were presented as pictures and the response required selection o f pictures. The performance o f the children with speech disorders was compared with that o f age-matched children who had no obvious speech impairment. Raine et al. (1991) found that the children with speech disorders had significantly slower speech than the group o f children without speech disorder, and they also had significantly shorter STM span. They interpret these findings to show that STM span is linked with speech rate. The group o f children showed a range o f different speech disorders. It is not clear from Raine et al.’s (1991) study what level o f speech output processing difficulty might be resulting in the children having a slow speech rate. Raine et al. (1991) observed less speech motor activity as the pictures were presented for recall, together with a reduced word length effect, for the children with speech disorders. This was interpreted to suggest that these children were less likely to use phonological coding to maintain the pictures in STM. The finding o f a reduced span in the children with speech disorders might therefore be the result o f use o f a different, less efficient, strategy for recall from that o f the children who did not have a speech disorder, rather than as a result o f slower speech rates.

This difference in strategy for remembering is similar to that found by Johnston and Anderson (1998). Children with poor reading skills showed a reduced word length effect for picture presentation o f items for recall, but not for verbal presentation. This suggested that this group o f children had also not used a phonological coding for the picture material. In this instance, Johnston and Anderson (1998) observed that the children with poor reading skills often showed lip movement when pictures were presented and the children reported that they named the pictures to themselves. They suggest that the children with poor reading skills may

have verbally coded the items on presentation, but have relied on visual information during recall. The observations o f lip movement and children’s report were not systematically measured, and so the researchers could not be certain that the children with poor reading skills had not differed from children with normal reading skills on these measures. In any case, it is clear that a different strategy was being used for STM.

C. Henry (1990) also found significantly slower speech rates and significantly reduced digit span in 3-5 year-old speech disordered children as compared to non speech-disordered. Speech rate was significantly correlated with span in children with and without speech disorders. Raine et al. (1991) examined the findings from a sub-group o f their children with speech disorders. The sub-group contained children who had dysarthria. This is a speech disorder o f neurological origin that affects motor execution skills directly. Again the dysarthric children showed significantly shorter spans as compared to children without speech disorders. This is in contrast to Bishop and Robson’s (1989) study which found that teenagers with anarthria or dysarthria performed in the same way as those with normal speech, in recalling word lists that were spoken or presented visually, with a picture-pointing response. These young people do not appear to have sufficient speech output skills for rehearsal. Bishop and Robson (1989) suggest that the rehearsal mechanism in STM must involve “abstract phonological representation, which, in normal individuals, would provide the input to a system that derives a speech motor program but does not itself contain an articulatory specification” (p. 139).

This finding was replicated by White, Craft, Hale and Park (1994), who also found no difference in span, despite significantly slower speech rates, in children with cerebral palsy, compared with children with no speech

difficulties. This was on a task with spoken presentation o f stimuli and spoken recall. The speech difficulties o f children with cerebral palsy are usually dysarthric in nature, i.e. a motor execution deficit. White et al. (1994) separate out overt and covert rehearsal, suggesting that central programming mechanisms are involved in both, with peripheral articulatory mechanisms implicated just in overt rehearsal. Their findings lead them to suggest that there are “different relationships between covert and overt rehearsal processes in some groups with impaired articulation” (p. 184). White et al. (1994) conclude that both groups were relying on covert rehearsal and that “normal speech rates are not necessary for development o f covert rehearsal” (p. 184).

Gathercole and Hitch’s (1993) hypothesis about the role o f the phonological readout process may account for these findings from Bishop and Robson (1989), and White et al. (1994). For the anarthric and dysarthric young people the impairment is in peripheral articulation (motor execution) and not in the more abstract phonological system. The phonological representations o f the items to be recalled were available for readout for the response, whether picture pointing or verbal.

Baddeley (1986) reports findings fi*om a number o f studies o f unimpaired STM span in adults with acquired anarthria or dysarthria, and concludes that peripheral speech is not essential for normal mechanisms o f STM. Rather, that the phonological loop and rehearsal processes o f the WM model are “apparently relying on central speech control codes which appear to function in the absence o f peripheral feedback” (p. 107). Within the context o f a speech processing model, this suggests that more central aspects o f speech output processing are implicated, such as access to motor programs, rather than motor execution skills.

Studies by Rochon, Caplan and Waters (1992) and Waters, Rochon and Caplan (1992) looked at STM in adults with acquired apraxia, a difficulty with motor programming and motor planning. These participants had reduced STM spans and slower speech rates when compared to adults without any speech difficulty. Analysis o f a range o f test and task results led Waters et al. (1992) to conclude that the rehearsal mechanisms involved in STM rely on speech planning processes that specify articulatory gestures rather than on the articulatory gestures themselves.

This evidence from children and adults with speech disorders links with more recent interpretations o f the evidence from experimental studies (Gathercole & Hitch, 1993, Henry 1994) to suggest that motor execution skills are not influential in STM. Other aspects o f speech processing may be implicated, for example access to motor programs and motor planning. Speech rate may be related to span because it is an observable measure that reflects these underlying, more central aspects o f speech output processing.

2.4 SPEECH INPUT SKILLS IN STM

Speech perception has been found to be significantly correlated with STM span in young adults (Watson & Miller, 1993). Research with deaf children suggests that there are deficits in performance on STM tasks (Campbell & Wright, 1990). It is possible that the development o f some aspect or aspects o f speech input processing skills are implicated in changes in STM with age, and in STM task performance.

2.4.! Identification Time Hypotheses

One hypothesis in accounting for differences in span has centred on the influence of identification time, i.e. the speed with which stimuli are

recognised and identified. This measurement is usually made by recording the time taken to repeat words, from the beginning o f the stimulus to the beginning o f the response. Case, Kurland and Goldberg (1982) found a significant correlation between mean identification time and span in 3 and 6 year-old children. This correlation was still significant when age was controlled for, suggesting that STM span could be accounted for by individual differences in identification time.

Case et al. (1982) recognised that this measure o f identification time included time spent in identifying words but also ‘preparation time to respond’. From a speech processing perspective, in addition to the levels of speech input processing required to access the appropriate phonological representation, such a measure will include the initial levels o f speech output processing o f time needed to access a motor program, and possibly preliminary stages of motor planning. Hulme and Tordoff (1989) evaluated Case et al.’s (1982) hypothesis about item identification, in the light of findings o f a relationship between speech rate and span. As Case et al.’s (1982) measure o f item identification included speech-onset time as well as item recognition, Hulme and Tordoff (1989) suggested that the item identification component may not be the critical determinant o f span. Rather the aspects o f speech output involved up to the point o f speech onset in the measure might have been the critical determinant.

Eleven year-old children were quicker to identify spoken and written words than 7 year-old children (Hitch et al., 1989b). However, neither identification time measure was significantly correlated with span. This was in accord with the findings o f Cohen, Quinton and Winder (1985, cited in Henry & Millar, 1993) that 10 year-old children were faster than 6 and 8

year-olds to identify spoken letters, masked within a string o f digits, but that this identification time measure did not correlate with span.

Hitch et al. (1993) found that item identification time was a poorer predictor o f span than speech rate. Although item identification time was significantly correlated with span, multiple regression analyses revealed that once the variance in STM span that was due to speech rate had been removed, there was no further significant variation due to item identification time. However, when item identification time was entered into the analysis first, significant further variation due to speech rate was found. Hitch et al. (1993) concluded that there was no support for the role o f identification time in the development o f STM span.

As in their analysis o f findings o f significant relationships between speech rate and span, discussed in the previous section, Henry and Millar (1991) argued that the use o f correlational evidence does not support a causal relationship between identification time and span. Henry and Millar found age-related increases in identification time for spoken words between 5 and 7 year-old children. However, when a group o f 5 year-olds and a group o f 7 year-olds were matched for identification time, STM span was still significantly better in the 7 year-old group. Henry and Millar (1991) suggested that identification time accounted for some but not all the variation in span and that the relationship between identification time and span does not reflect causation.

Following a review o f the literature, Henry and Millar (1993) conclude that there is insufficient support for the hypothesis that identification time is responsible for developmental changes in span, but acknowledge that it may be “a subsidiary factor in span growth” (p. 247).