The role of short-term memory impairment in nonword repetition, real word repetition, and nonword decoding: A case study

Abstract

In a companion study, adults with dyslexia and adults with a probable history of childhood apraxia of speech showed evidence of difficulty with processing sequential information during nonword repetition, multisyllabic real word repetition and nonword decoding. Results suggested that some errors arose in visual encoding during nonword reading, all levels of processing but especially short-term memory storage/retrieval during nonword repetition, and motor planning and programming during complex real word repetition. To further investigate the role of short-term memory, a participant with short-term memory impairment (MI) was recruited. MI was confirmed with poor performance during a sentence repetition and three nonword repetition tasks, all of which have a high short-term memory load, whereas typical performance was observed during tests of reading, spelling, and static verbal knowledge, all with low short-term memory loads. Experimental results show error-free performance during multisyllabic real word repetition but high counts of sequence errors, especially migrations and assimilations, during nonword repetition, supporting short-term memory as a locus of sequential processing deficit during nonword repetition. Results are also consistent with the hypothesis that during complex real word repetition, shortterm memory is bypassed as the word is recognized and retrieved from long-term memory prior to producing the word.

In a companion study (Peter, Lancaster, Vose, Middleton, & Stoel-Gammon, 2017), the hypothesis was evaluated that a domain-general sequential processing deficit is shared by individuals with dyslexia (DYX) and individuals with childhood apraxia of speech (CAS). Participants were 22 adults with DYX, 10 adults with a probable history of CAS (phCAS) and 22 typical controls (CTR). They completed three different linguistic tasks, nonword repetition, multisyllabic real word repetition, and nonword decoding. The two disorder groups showed more sequencing and substitution errors than the controls in all tasks but an especially disproportionally large number of sequencing errors during the nonword repetition task. In both groups, loss of serial order likely occurred at the level of visual encoding during nonword decoding, multiple levels but especially short-term memory during nonword imitation, and motor planning/programming during multisyllabic real word repetition.

The purpose of the present case study was to further investigate the role of short-term memory in the three tasks. The rationale was that newly encoded phoneme sequences in nonword repetitions, presented only once and fleetingly in the auditory modality, fade rapidly from short-term memory. Phonemes that were retained in short-term memory may undergo sequential rearrangement and phonemes that were not retained at all may be substituted with other sounds during speech production. In contrast, during nonword decoding, the grapheme sequence remains stable and visible for several seconds, allowing for refreshing the memory buffer as necessary. During real word imitation, the spoken target is recognized and retrieved from long-term memory, bypassing the short-term memory level. The levels of phonological assembly, phonological output buffer, motor planning/programming and motor execution are common to all three tasks. The hypothesis is therefore that persons with impaired short-term memory will produce more sequence as well as substitution errors than typical controls during nonword repetition but not during real word repetition and nonword decoding.

Method

Towards investigating the role of memory in the three word-level tasks, a participant with short-term memory impairment (MI) was recruited. This participant, a female aged 65 years, self-reported short-term memory difficulties, and family members confirmed these difficulties. The cause of her MI could not be ascertained.

For purposes of validating the MI, the participant with MI was asked to complete not only the same battery of tasks as the CTR, phCAS and DYX groups described in the companion study but also a set of additional tasks with a high short-term memory loads. Like the CTR, phCAS and DYX groups, she completed untimed tests of word and nonword reading (Word Identification, WID and Word Attack, WATT, subtests of the Woodcock Reading Mastery Tests – Revised (WRMT-R) (Woodcock, McGrew, & Mather, 2001)). The Sight Word Efficiency (SWE) subtest and the Phonemic Decoding Efficiency subtest (PDE) of the Test of Word Reading Efficiency (Torgesen, Wagner, & Rashotte, 2012) assess word and nonword reading ability, respectively, under timed conditions. The Spelling subtest from the Wechsler Individual Achievement Test – II (WIAT-II) (Wechsler, 2005) tests sight word spelling ability. Diadochokinetic testing included rapid repetitions of the monosyllable /pa/ and the disyllable /pata/. Syllable durations were converted into z scores using published norms (Fletcher, 1972). The participant also completed the two verbal processing tasks from the Reynolds Intellectual Assessment Scales (RIAS) (Reynolds & Kamphaus, 2003). In one task, the participant is asked to complete a verbal analogy and in the other, to provide a term that corresponds to a definition. The Verbal Intelligence Index (VIX) is computed based on the average performance on these tasks. Nonword Repetition (NWR) subtest from the Comprehensive Test of Phonological Processing (CTOPP) (Wagner, Torgesen, & Rashotte, 1999). Of these tasks, only the NWR taxes short-term memory.

In addition, the participant completed several tasks with high short-term memory loads. The Recalling Sentences subtest from the Clinical Evaluation of Language Fundamentals – 4 (CELF-4) (Semel, Wiig, & Secord, 2003) requires listening to sentences presented one at a time, then repeating them verbatim. According to the test publishers, the Recalling Sentences subtest addresses syntactic complexity, word length and idea density in the verbal information to be remembered. The participant with MI also completed two additional nonword repetition tasks, the Nonword Repetition Task (NRT) (Dollaghan & Campbell, 1998) and the Syllable Repetition Test (SRT) (Shriberg et al., 2009). Qualitative error type analysis was only completed for the NRT because of the greater variety of consonants and vowels, compared to the SRT, whereas z scores were calculated for both tests based on published norms (http://www.waisman.wisc.edu/phonology/techreports/TREP14.pdf).

The participant’s low scores in all tasks with high short-term memory loads and her typical performance on all but one task without such loads are consistent with the short-term memory deficit (Table 1). The RIAS-VIX probes static knowledge about word meaning and semantic relationships stored in long-term memory, whereas the Recalling Sentences subtest requires repeating verbally presented sentences that require storing in, and retrieval from, short-term memory. The unexpectedly low score in the TOWRE-PDE subtest, designed to measure nonword decoding ability under a timed condition, may reflect an overall decrease in processing speed or slow motor speed, either of which is also consistent with the slow syllable repetitions during both monosyllabic and multisyllabic DDK tasks. Further evidence for this hypothesis is provided by performance within normal limits during the WATT subtest that measures nonword decoding under an untimed condition.

Table 1.

Descriptive measures for the participant with short-term memory impairment.

Working Memory Load	Test	Standard Score	Z Score
Low	RIAS VIXa	103	0.20
	DDK -/papa/b	N/A	−1.71
	DDK –/pata/b	N/A	−2.08
	WIDa	97	−0.20
	WATTa	92	−0.53
	TOWRE - SWEa	90	−0.67
	TOWRE - PDEa	82	−1.20
	WIAT II Spellinga	114	0.93
High	CELF-4 Recalling Sentencesc	5	−1.67
	NWRb	6	−1.33
	SRT	N/A	−9.07
	NRT	N/A	−2.31

Notes.

^aMean = 100, standard deviation = 15;

^bMean = 0, standard deviation = 1;

^cMean = 10, standard deviation = 3; RIAS VIX = Reynolds Intellectual Assessment Scales Verbal index; DDK = diadochokinetics; WID = Word Identification; WATT = Word Attack; TOWRE – SWE = Test of Word Reading Efficiency: Sight Word Efficiency; TOWRE – PDE = Test of Word Reading Efficiency: Phonemic Decoding Efficiency; WIAT II = Wechsler Individual Achievement Test – II; CELF-4: Clinical Evaluation of Language Fundamentals, Fourth Edition; NWR = Nonword Repetition; SRT = Syllable Repetition Test; NRT = Nonword Repetition Task.

As in previous studies (Button, Peter, Stoel-Gammon, & Raskind, 2013; Peter, Button, Stoel-Gammon, Chapman, & Raskind, 2013) and the companion study (Peter et al., 2017), errors were analysed with phonological process analysis, which is typically used to identify patterns in speech sound errors produced by children (Edwards & Shriberg, 1983; Stoel-Gammon & Dunn, 1985; Stoel-Gammon & Stone, 1991). Errors were classified as sequential errors (assimilations, migrations, metatheses, omissions and insertions), where the serial order of the target was re-arranged, or substitution errors where the identity of the target element was obliterated. The participant’s errors in each class and type during the three word-level tasks (nonword repetition, real word repetition and nonword decoding) were descriptively compared to the median counts in the three participant groups (CTR, DYX, phCAS), as z tests were not appropriate given the non-homogeneity of the error sums.

Results

During the NWR task, the participant with MI produced 18 sequencing and 8 substitution errors (Table 2). The number of sequencing errors was most similar to that in the phCAS group and the number of substitution errors was the same as in the two disorder groups. Assimilation was the most commonly occurring sequencing error type in the DYX and phCAS groups. The participant with MI produced similar numbers of assimilations but nearly half (8/18) of her errors were migration errors. The error profile during the second nonword repetition task, the NRT, showed 15 sequencing and 8 substitution errors. During this task as well, migration errors accounted for approximately half (8/15) of all sequencing errors and one-third (5/15) were assimilation errors.

Table 2.

Experimental measures obtained by the participant with memory impairment by error class, error type and task.

	Sequence							Identity
Task	Assimilation (Ant.)	Assimilation (Pers.)	Migration	Metathesis	Omission	Insertion	Sum	Substitution
NWR	2	3	8	1	1	3	18	8
NRT	2	3	8	1	1	0	15	8
MSW	0	0	0	0	0	0	0	0
WATT	0	1	0	1	5	1	8	1

Notes. NWR = Comprehensive Test of Phonological Processing - Nonword Repetition subtest, NRT = Nonword Repetition Test, MSW = Catts’ (1986) Multisyllabic Word repetition task, WATT = Woodcock Johnson Tests of Achievement - Word Attack subtest.

During the multisyllabic real word repetition task, the participant with MI did not produce any errors. This result was the same as the CTR group.

During the nonword decoding task, the participant produced 8 sequencing errors, which was most similar to the median error count in the two disorder groups. As in these two groups, omission errors were the most prevalent sequencing error type. The participant produced only one substitution error. The lone substitution error consisted of rendering the letter ‘n’ as [m].

Discussion

The fact that the participant with MI had no difficulty imitating multisyllabic real words is consistent with an intact speech production system downstream of the phonological assembly. Regarding errors during nonword repetition using the NWR, this participant’s number of sequencing errors was greater than that in the control group but equivalent to that in the DYX and phCAS groups, indicating difficulty with sequential information. Her high number of substitution errors reflects loss of identity of the sounds in the sequence. The distribution of sequencing errors during the NWR was replicated closely during the NRT, supporting the validity of the NWR results. During nonword decoding, the number of sequencing errors was higher than expected, consisting mostly of omissions. This error type was also most prevalent in the DYX and phCAS groups. Similar to these two groups, it is possible that the omissions represent deficits at the level of visual encoding, not short-term memory, as her omission of the letter ‘r’ in the nonword ‘byrcal’ resulted in a phoneme change from /ɝ/ to [aɪ]. In contrast, there was only one substitution error and it may have been triggered by the visual similarities between the letters ‘n’ and ‘m’.

Many sequencing errors during the NWR were also seen in the DYX and phCAS groups. In these two groups, similar to the participant with MI, sequencing errors may have arisen mainly due to constraints at the level of short-term memory, although this task also taxes encoding and motor transcoding processes.

The fact that the participant with MI produced high numbers of substitution errors during nonword repetition but not during nonword decoding is consistent with impaired short-term memory as the locus of impairment for this task, where, in the absence of a visual representation of the target, a memory of a placeholder for some sounds was retained but not the sounds themselves. During this task, she also produced many sequencing errors, consistent with a memory of the identity of some sounds but not their position in the word.

Finally, the error-free repetitions of multisyllabic real words in the presence of high sequencing and substitution error counts during nonword repetition confirms the hypothesis that during real word repetitions, encoding into short-term memory is bypassed and replaced by word recognition and retrieval from long-term memory.