Académique Documents
Professionnel Documents
Culture Documents
Patients with semantic STM deficits have difficulty comprehending sentences that require the retention of
several lexical-semantic representations prior to their integration into higher-level propositions (Martin,
1995; Martin & Romani, 1994). In Experiment 1, patients with a semantic retention deficit had difficulty
with the same type of constructions in speech production, namely noun phrases with one or two pre-
nominal adjectives. Their performance improved when they could produce the nouns and adjectives in
sentence form, which placed smaller demands on lexical-semantic retention. In Experiment 2 these
patients were better able to produce syntactically complex sentences than the prenominal adjective
phrases having an equal number of content words, indicating that the findings in Experiment 1 could not
be attributed to syntactic complexity. These patients produced more pauses in the sentence constructions
in Experiments 1 and 2, suggesting that the timing of such productions is abnormal. In contrast, patient
EA, with a phonological retention deficit, performed better than the patients with a semantic retention
deficit on the AN phrases despite having a smaller STM span. She showed no significant benefit of
producing sentence compared to phrase constructions, and also made fewer and shorter pauses than the
other patients. These findings support the multiple capacities view of verbal working memory and suggest
that the same semantic retention capacity used in language comprehension is used in speech production.
Requests for reprints should be sent to Randi C. Martin, Department of Psychology, Rice University, P.O. Box 1892, Houston, TX
77251, USA. Email: rmartin@rice.edu
This research was supported by NIH grant DC 00218 to Randi C. Martin at Rice University. The authors would like to thank Mike
Katz, Ann-Marie Lobo, and Frank Tamborello for assistance in patient testing and data analysis.
1994; Dell, 1986). At the highest level, the mes- there was no phonological effect for words related
sage to be expressed is represented in a nonverbal to the second noun. Given this evidence of a
format. The message is used to select lexical greater scope of planning at the semantic level, it
representations and a syntactic structure to is possible that there are greater capacity demands
express the relations among the lexical repre- at the level involved in maintaining lemma
sentations. It should be noted that the lexical representations.
representations at this level are presumed to be
representations of semantic and syntactic infor-
mation (i.e., ‘‘lemmas’’; Kempen & Huijbers, Components of verbal short-term
1983) rather than phonological representations. memory and their relation to
Some recent speech production models have sentence planning
implied that lemmas contain purely syntactic
representations (Bock & Levelt, 1994; Levelt, Recent evidence suggests that a number of dif-
Roelofs, & Meyer, 1999). However, lemmas are ferent types of representations are involved in the
portrayed in graphical depictions of these models maintenance of words during standard memory
as being essentially empty nodes that link con- span tasks. A large body of research implicates a
ceptual representations to syntactic information. role for phonological representations, and many
In this paper we will use ‘‘lemma’’ to refer to this models of short-term memory have focused on the
type of representation that provides a means of retention of phonological codes (Baddeley, 1986;
connecting conceptual and syntactic specifications Schweickert, Guentert, & Hersberger, 1990).
of words. Further downstream in the planning However, other studies involving normal subjects
process, the phonological representations for the and brain-damaged patients indicate that seman-
words are retrieved. In these models, information tic representations are involved as well. For
may flow from one level to the next before a example, studies of normal subjects have shown
complete representation of the entire utterance is that span is greater for words than nonwords
constructed at the prior level. For example, (Crowder, 1978; Hulme, Maugham, & Brown,
phonological encoding of the first noun phrase 1991; Schweickert, 1993), greater for words from
may proceed while construction of the syntactic the same category than for words from different
representation of the second noun phrase is in categories (Poirier & St. Aubin, 1995), and greater
progress. for high than low imageability words (Bourassa &
The scope of planning, in terms of the propor- Besner, 1994). Also, span for foreign words
tion of the message being planned, appears to vary increases when subjects learn the meanings of the
depending on the level at which the planning is foreign words (Hulme et al., 1991). Neuro-
taking place. At the message level, the entire psychological evidence comes from the work of
thought may be represented. However, at sub- Patterson and colleagues and of N. Martin and
sequent levels, smaller portions may be planned. Saffran. Patterson, Graham, and Hodges (1994)
Data from naturally occurring speech errors and and Knott, Patterson, and Hodges (1997)
from experimental studies suggest that the scope demonstrated that patients with dementia that
of planning is greater at the level of lemma affects semantic representations show a larger
selection than at the phonological level (Garrett, word span for known words than for unknown
1980; Meyer, 1996; Smith & Wheeldon, 1999). For words (that is, words for which they no longer
instance, whole word exchange errors (e.g., ‘‘I left know the meaning). Saffran and N. Martin (1990;
the cigar in my briefcase’’ ? ‘‘I left the briefcase Martin & Saffran, 1992, 1997) analysed the word-
in my cigar’’) tend to occur across a larger number processing and short-term memory deficits of
of intervening words than do sound exchange patients with phonological vs semantic processing
errors (e.g., ‘‘strictly speaking’’ ? ‘‘spictly deficits and found distinctive short-term memory
streaking’’) (Garrett, 1980). Also, in a picture patterns with regard to recency and primary
description paradigm with spoken word dis- effects in serial recall and the effects of image-
tractors, Meyer (1996) found that onset latencies ability and frequency at early vs late serial posi-
for a phrase such as ‘‘the boat and the car’’ were tions. That is, patients with phonological
delayed by the presentation of a word semanti- processing deficits tend to show greater primacy
cally related to either the first or second noun. than recency effects and greater effects of
Facilitation in onset latencies was observed for imageability at late serial positions, whereas
words phonologically related to the first noun but patients with semantic processing deficits tend to
SEMANTIC RETENTION AND SPEECH PRODUCTION 263
show greater recency than primacy effects and show poor performance on list recall tasks that
large effects of imageability at early list positions. require phonological output, but good perfor-
Thus, there is evidence from both normal subjects mance on recognition tasks tapping input reten-
and brain-damaged patients supporting the tion, such as digit matching span tasks or probe
involvement of multiple linguistic codes in verbal tasks (Allport, 1984; Romani, 1992; Shallice,
short-term memory. Rumiati & Zadini, 2000). The opposite dissocia-
Neuropsychological data have also indicated tion of an impaired phonological input buffer but
that the phonological and semantic retention preserved output buffer is demonstrated by
capacities may be selectively affected by brain patients who have very reduced spans and poor
damage, independently of damage to phonologi- performance on input tasks like those just men-
cal and semantic processing abilities. Martin, tioned, but who produce speech that is within
Shelton, and Yaffee (1994) reported two patients normal range in terms of sentence length, gram-
who both had very reduced spans, but who showed matical complexity, and rate of pausing and hesi-
different effects of phonological and semantic tations (R. Martin et al., 1994; Shallice &
variables on span. Both patients showed normal or Butterworth, 1977). Under the assumption that an
near normal levels of single word processing in output phonological buffer is needed to plan
both comprehension and production tasks. Patient speech production, the good language production
EA demonstrated reduced phonological effects, for these patients suggests that the output buffer is
but preserved semantic effects on memory span. preserved despite the severe restriction in the
For example, she showed no phonological simi- input buffer.
larity effect in the visual modality and no word Patient MS (Martin, Lesch, & Bartha, 1999)
length effect in either auditory or visual modali- provides another source of evidence for the
ties, and showed better performance with visual separation of input and output buffers. We have
than auditory presentation (the reverse of the hypothesised that MS has difficulty, not with the
normal pattern). However, she showed a large output buffer per se, but rather with activating
advantage for words over nonwords, suggesting output phonological representations from lexical-
that the semantic information in the words aided semantic representations. His picture naming is
her performance. Two short-term memory probe severely impaired, but when he is unable to pro-
tasks were administered, a rhyme probe task that duce a picture name, he typically produces a long
tapped phonological retention and a category description of the object, indicating preserved
probe task that tapped semantic retention. EA semantic knowledge. This patient has unimpaired
performed slightly better on the category probe speech perception but has difficulty specifically in
than the rhyme probe task, whereas normal sub- retrieving output phonological representations.
jects showed a substantial advantage on the rhyme He performed normally on nonword list recall,
probe task. In contrast to EA, patient AB showed suggesting that both his input and output buffers
normal phonological effects in span tasks but no are preserved and that he can translate directly
advantage for words over nonwords. AB per- between input and output phonological repre-
formed worse than EA on the category probe task, sentations. On short-term memory tasks involving
but better than EA on the rhyme probe task. Thus only input phonological representations for words
across the two patients there was a double dis- (such as rhyme probe or order probe tasks), he
sociation between deficits in semantic vs phono- performs at a normal level. However, when
logical retention. This dissociation has recently required to reproduce a list of words, his perfor-
been replicated in a study of two children who had mance was impaired, particularly for low-
sustained severe closed head injury (Hanten & frequency, low-imageability words. He would
Martin, 2000). sometimes provide semantic descriptions of the
In addition to the distinction between semantic words in the list. His performance suggests that
and phonological retention capacities, there is the connections used for activating output
evidence that a further distinction needs to be phonological representations from lexical-
made between the capacities involved in the semantic representations are damaged, which has
retention of input and output phonological codes a greater impact on his production of low- than
(Allport, 1984; Romani, 1992; Shallice & Butter- high-frequency words in both naming and list
worth, 1977; but see N. Martin & Saffran, 1992). recall. Thus, he does not show the boost in per-
Evidence for a selective impairment of a phono- formance for words compared to nonwords that is
logical output buffer is provided by patients who shown by normal subjects because normal subjects
264 MARTIN AND FREEDMAN
receive activation both from direct translation in limited-capacity buffers specific to each level
from the input phonological representations and of representation.
from activation flowing from lexical-semantic The lexical knowledge network represented on
representations. the left-hand side of Figure 1 is similar to that
Based on the findings of semantic and phono- proposed by Dell and colleagues in their model of
logical contributions to memory span, we have word production (Dell, 1986; Dell & O’Seaghdha,
developed a model of short-term memory in 1992; Dell et al., 1997). In Dell’s (1986) model of
which there is a close connection between levels sentence production, the different levels of
of representation in language processing and representation in the lexical network are tied to
those involved in short-term memory. As shown slots in syntactic, morphological, and phonological
in Figure 1 (Martin et al., 1999), language pro- planning frames. It seems plausible to assume that
cessing is supported by interactive activation of the phonological buffer that we have hypothesised
representations in the long-term knowledge is identical to the phonological planning frame in
store. During word recognition, acoustic infor- Dell’s model. The relation between the lexical-
mation is mapped onto a phonological form, semantic buffer in Figure 1 and Dell’s planning
which then accesses a lexical form, which then frames could take different forms. One possibility
activates a semantic representation. In word pro- is that the lexical-semantic buffer is used to hold
duction, activation begins at the semantic level, semantic and syntactic specifications of several
which then activates a lexical representation, words simultaneously and then these representa-
which then activates its phonological representa- tions are linked to a separate syntactic frame. A
tion, which is used to guide articulation. The second possibility is that the lexical-semantic
output from language processing is maintained buffer is equivalent to the syntactic planning
Figure 1. Model of language processing and short-term memory (based on Martin et al., 1999).
SEMANTIC RETENTION AND SPEECH PRODUCTION 265
frame. If so, then to account for the patients’ (1994) and Martin (1995) demonstrated that
short-term memory deficits, one would have to patients who show semantic retention deficits on
assume that a syntactic frame was involved in memory span tasks perform poorly on sentence
serial list recall. As Barnard (1985) suggested, one comprehension when they have to maintain sev-
might assume that the syntactic frame for a word eral lexical-semantic representations prior to their
list was quite simple and similar to that involved in integration into higher-level propositional repre-
representation of a compound noun phrase (e.g., sentations. These studies employed an auditory
‘‘Jeeps, cars and buses . . .’’). In either case, we sentence anomaly judgement task, manipulating
would assume that links between the lexical- the number of adjectives before or after a noun or
semantic buffer or syntactic frame to the repre- the number of nouns before or after a verb. For
sentations in the knowledge store would serve to example, the ‘‘before’’ condition with adjectives
keep these representations active, and damage to used anomalous sentences such as ‘‘The rusty, old,
the buffer or to the links should have negative red swimsuit . . .’’ or ‘‘The rusty swimsuit . . .’’. The
consequences for production. We prefer the for- ‘‘after’’ condition used anomalous sentences of
mer hypothesis that there is a separate lexical- the form ‘‘The swimsuit was old, red, and
semantic retention buffer that feeds into a syn- rusty . . .’’ or ‘‘The swimsuit was rusty . . .’’.
tactic frame, rather than equating the lexical- Matching sensible sentences were also presented
semantic buffer with the frame. It seems easier to (for example, ‘‘The rusty pail . . .’’). A similar
account for the patients’ span deficits in terms of manipulation was used for sentences that varied
overly active decay or increased interference the number of nouns that came before or after a
among lexical-semantic representations. If the verb. For example, a ‘‘before’’ sentence with three
buffer were equated with the syntactic frame, one nouns was ‘‘The cloth, the vase and the mirror
would have to assume some deficit in the capacity cracked . . .’’ and the matching ‘‘after’’ sentence
of the syntactic frame itself or possibly in the links was ‘‘The movers cracked the mirror, the vase and
between the frame and the knowledge store. The the cloth . . .’’. It was reasoned that the ‘‘before’’
results from Experiment 2 provide some data on conditions should make a larger demand on
this issue. lexical-semantic retention because integration of
lexical-semantic information was delayed. In the
‘‘after’’ conditions, integration was immediate.
Relation between capacities involved That is, the meaning of an adjective had to be
in language production and maintained until it could be integrated with a
language comprehension noun, and the meaning of a noun had to be
maintained until it could be integrated in terms of
As discussed earlier, evidence suggests that there its thematic role with respect to the verb.
are separate buffers involved in retaining input Martin and Romani (1994) and Martin (1995)
phonological forms and output phonological showed that two patients (AB and ML) with
forms. This separation raises the question of semantic STM deficits performed very poorly in
whether there are separate lexical-semantic the ‘‘before’’ condition when two or three words
retention capacities involved in comprehension had to be maintained prior to integration. They
and production. Kempen and Hoenkamp (1987) made a large number of errors and showed a very
suggested that the same syntactic procedures are large effect of the number of nouns or adjectives
involved in production and comprehension and that had to be maintained prior to integration. In
recent empirical evidence supports this view fact, both performed near chance level with two or
(Branigan et al., 1995). However, as indicated three adjectives or nouns in the ‘‘before’’ condi-
earlier, it is possible that the lexical-semantic tion (about 37–40% errors for each patient in
retention buffer is separate from the syntactic these conditions where 50% would be chance),
frame, and thus other evidence would be needed compared with 10–15% errors with one preceding
to address whether this buffer is shared in com- adjective or noun. In contrast, in the ‘‘after’’
prehension and production. In a number of pre- condition, they made many fewer errors overall
vious studies, we have demonstrated that patients and showed little effect of the number of adjec-
with a semantic STM deficit have difficulty with tives or nouns. In contrast, patient EA, with a
sentences with a high semantic load (Martin & phonological retention deficit, performed like
Feher, 1990; Martin & Romani, 1994; Martin et normal subjects with slightly worse performance
al., 1994). More specifically, Martin and Romani overall in the ‘‘before’’ condition. However, she
266 MARTIN AND FREEDMAN
did not show the interaction between the ‘‘before/ Freedman, 2001). Thus, in an adjective-noun or
after’’ manipulation and distance that was shown adjective-adjective-nou n phrase, where the head
by the semantic retention deficit patients. of the phrase is the noun, the noun plus all pre-
If the same lexical-semantic retention capa- ceding adjectives must be planned at the lexical-
cities are used in both comprehension and pro- semantic level prior to phonological encoding of
duction, then patients with semantic retention the phrase. Patients who are unable to maintain all
deficits might have difficulty producing the same of these lexical-semantic representations simul-
types of constructions that caused difficulty in taneously might either fail entirely at producing
comprehension (namely, noun phrases with sev- the phrase, or produce it in a piecemeal fashion
eral prenominal adjectives and subject noun with pauses intervening between each content
phrases with more than one noun). Recent models word.
of language production have assumed that plan- Some evidence from normal subjects is con-
ning proceeds incrementally (de Smedt, 1996; sistent with this lexical head principle. Schriefers
Kempen & Hoenkamp, 1987; LaPointe & Dell, (1993) examined the effects of visually presented
1989; Smith & Wheeldon, 1999). Whereas a distractor words on the production of adjective-
message-level representation may be developed noun phrases (in Dutch) describing coloured
that spans an entire clause or more, planning at objects (e.g., ‘‘groen bed’’ ? ‘‘green bed’’). Dis-
the syntactic level proceeds in a phrase-by-phrase tractor words semantically related to either the
fashion in terms of both development of gram- noun or the adjective delayed onset latencies
matical structure and the attachment of lexical- relative to unrelated distractors, suggesting that
semantic representations to this structure. If we subjects were planning both words at the semantic
assume that all the lexical-semantic representa- level prior to speech onset. However, this effect
tions in a phrase must be planned simultaneously was obtained only for phrases without a deter-
prior to the initiation of phonological encoding of miner. For phrases with a determiner (‘‘het groene
the phrase, then patients with a lexical-semantic bed’’ ? ‘‘the green bed’’), interference was
retention deficit should have difficulty producing obtained only for distractor words semantically
phrases containing several content words. A related to the noun. This latter result might sug-
phonological retention deficit that is limited to an gest that only the lexical head needs to be planned
input buffer deficit would be predicted to have no at the semantic level and not the preceding
effect on production, as only the output phono- adjective. In Dutch, however, when both a deter-
logical buffer should be involved in planning miner and adjective are present, only the deter-
phonological representations for speech produc- miner agrees in gender with the noun. When no
tion (Martin & Freedman, 2001; Martin et al., determiner is present, the adjective agrees in
1999; Shallice & Butterworth, 1977). gender. Thus, for the phrases with determiners, it
It might be argued that even though normal is possible that determiner selection was slow
subjects might plan all of the lexical-semantic relative to adjective selection, with the result that
representations for a phrase simultaneously, there interfering effects of the semantically related
may be no necessity to do so. That is, if patients adjective distractor were hidden by the relatively
can keep the message to be expressed in mind, long time needed to select the correct determiner.
they might be able to retrieve the lexical-semantic Other results that are somewhat relevant to
representations for each word sequentially—pro- the lexical head principle come from a recent
vided that they can maintain the syntactic struc- study by Smith and Wheeldon (1999). They con-
ture for the phrase in order for the words to be trasted onset latencies to describe moving pic-
produced in the right order. However, it is pos- tures with sentences of two types: (1) simple/
sible that in order to produce speech fluently with complex: The ball moves above the faucet and
normal intonational contours, several lexical- the sock; (2) complex/simple: The ball and the
semantic representations must be planned and faucet move above the sock. Even though the
maintained in parallel while phonological encod- two sentence types were matched in length and
ing is carried out. In the present paper, we inves- content words, initiation times were longer for
tigated what we have termed the ‘‘lexical head the complex/simple sentences than the simple/
principle’’, that is, that speakers must plan the complex sentences. The results suggest that
lexical-semantic representation for the head of the speakers were planning both nouns in the sub-
phrase and that for all of the content words prior ject noun phrase at least at the semantic level in
to that head within the same phrase (Martin & the complex/simple sentences prior to initiating
SEMANTIC RETENTION AND SPEECH PRODUCTION 267
articulation. However, the use of a compound phrases, pictures of three different objects (hair,
noun phrase somewhat complicates the issue of leaf, and curtain) were constructed, which varied
what constitutes a phrase. The compound noun on three different dimensions (e.g., Hair—long vs
phrase consists of two noun phrases, each of short, blonde vs black, straight vs curly). The AN
which has its own head. Thus, if subjects were phrases were elicited by showing two pictures side
only planning the first noun phrase of the com- by side, which were identical except for one rele-
pound noun phrase, then no effect of complexity vant dimension (e.g., long straight black hair vs
should have been obtained. The results suggest short straight black hair). The target picture was
instead that subjects planned the entire subject highlighted to indicate which picture should be
noun phrase. described. Subjects were instructed to produce an
A preliminary study from our lab (Martin & AN phrase to describe the highlighted picture
Freedman, 2001; Martin & Romani, 1995) sup- which would differentiate it from the other pic-
ported the contention that patients with a ture. The AAN phrases (e.g., ‘‘long curly hair’’)
semantic retention deficit would have difficulty were elicited by showing three pictures. The target
producing utterances that, according to the lexical picture was highlighted, and the other pictures
head principle, would require the maintenance of differed from the target on only one of the rele-
two or more lexical-semantic representations. vant dimensions (e.g., long curly hair [target]
Patients were asked to describe pictures using shown with long straight hair and short curly hair;
simple adjective-noun (AN) and adjective- see Figure 2).
adjective-noun (AAN) phrases. Patient EA, with As shown in Table 1, patient EA with a
an input phonological retention deficit, was pre- phonological STM deficit performed normally on
dicted to have no difficulty with this task. Patients both the one- and two-adjective conditions in the
ML and AB, who had lexical-semantic retention noun phrase production task. The two patients
deficits and who had done poorly on compre- with semantic retention deficits were able to name
hending sentences with several prenominal the single nouns and adjectives perfectly, but
adjectives, were predicted to have difficulty, par- performed very poorly on the AN and AAN
ticularly with the AAN phrases. To elicit the combinations, well below the range of controls.
1
ML performed substantially better on the AN phrases and Phonological STM Deficit
somewhat better on AAN phrases in Experiment 1 than he did EA
in the preliminary study, even though his production of the Time to onset 1.4 1.0
single adjectives and nouns was slightly better in the pre- Mean # pauses 0.3 0.6
liminary study. In the earlier study, he self-corrected to the Mean pause length 1.3 1.3
appropriate response on 60% of the trials in the AN condition
and 30% of the trials in the AAN condition. In Experiment 1,
he self-corrected to the correct response on 22% of the trials in
both conditions. His accuracy on final responses was similar
across the two experiments. A possible explanation of the GR showed a large effect of phrase vs sentence
differing results is that ML delayed his onset of the adjective- construction on percent correct (improving 33%
noun phrases in Experiment 1 more so than in the preliminary in the one-adjective condition and 16% in the two-
study in order that he could self-correct his utterance internally adjective condition). Collapsed across the one-
before beginning speaking. In the preliminary study, he may
have begun speaking quickly and then self-corrected overtly to
and two-adjective conditions, he was significantly
the correct utterance. However, we do not have latency data more accurate in the sentence than in the phrase
from the preliminary study to test this hypothesis. versions of the task, w2 = 4.5, p = .03. GR showed
272 MARTIN AND FREEDMAN
only relatively small effects in response onset. For correct for describing the picture, but were not the
the one-adjective condition, GR was 0.5 seconds appropriate ones for distinguishing the target
faster in the sentence than the phrase construc- picture from the comparison picture(s). On the
tions, t(21) = .83, p = .21, and for the two-adjective two-adjective trials (averaging across phrase and
condition, he was 0.6 seconds faster producing the sentence constructions), EA produced incorrect
sentence construction, t(11) = .50, p = .31. Like adjectives on 17% of the trials, whereas ML did
ML, GR also showed an increase in the mean this on 42% and GR on 61% of the trials. Thus, it
number of pauses within the response for the appears that ML and GR had a great deal of dif-
sentence compared to the phrase productions, ficulty suppressing the irrelevant information in
increasing from 0.2 to 0.9 pauses in the one- the pictures for the two-adjective condition,
adjective condition, t(21) = 2.6, p = .008 and from although EA also had some difficulty. ‘‘Opposite
1.0 to 2.0 in the two-adjective condition, t(11) = adjective’’ refers to producing a phrase with an
2.06, p = .03). adjective opposite to the target adjective (e.g.,
EA showed a 6% decline in performance from ‘‘short hair’’ for the target ‘‘long hair’’). ‘‘Omit
the phrase to the sentence constructions for the adjective’’ refers to producing a phrase but leaving
one-adjective condition, but an improvement of out one of the target adjectives (e.g., ‘‘blonde
22% for the two-adjective condition. The differ- hair’’ for the target ‘‘long blonde hair’’). GR was
ence between the phrase and sentence construc- the only patient to commit these two error types.
tions was non-significant for the combined data Finally, EA used incorrect verb agreement on two
(78% vs 86% correct, w2 = .84, p = .36). EA sentence trials (e.g., ‘‘The curtain are red’’).
showed a small advantage in onset times in the
sentence constructions that was similar to that of
GR. However, the differences were highly sig- Discussion
nificant for EA, due to her much lower variance in
We had predicted that ML and GR would be better
onset latencies. For the one-adjective condition,
able to produce the sentence form than the phrases,
she was 0.5 seconds faster for the sentence than
whereas this manipulation should have little effect
the phrase version, t(20) = 4.9, p < .001. For the
for EA. These predictions were borne out for the
two-adjective condition, she was 0.4 seconds fas-
most part. GR, but not ML or EA, showed sig-
ter, t(26) = 3.9, p < .001. Although both EA and
nificantly greater accuracy in the sentence con-
ML showed a significant effect of construction
structions. ML showed substantially faster onset
type on onset latency, the effect was clearly much
latencies for correct initial responses in the sen-
larger for ML. A two (patient) 6 two (condition)
tence than in the phrase constructions. Although
ANOVA on the combined data from the one- and
EA also showed significantly faster onset latencies
two-adjective conditions confirmed that there was
in the sentence constructions, the effect for ML was
a significant interaction between patient and
about 20 times as large. (ML was 8.0–12.0 seconds
condition, F(1, 94) = 17.9, p < .001. Furthermore,
faster in the sentence than the phrase condition,
EA demonstrated many fewer pauses in all con-
whereas EA was 0.4–0.5 seconds faster).2 This
struction types, and a smaller increase in number
difference between GR and ML, in that GR shows
of pauses between the phrase and sentence con-
a large benefit of construction type in his accuracy
structions. For the one-adjective condition, she
while ML shows most of the benefit in his time to
increased from 0.0 pauses in the phrase construc-
tions to 0.1 pauses in the sentence constructions,
t(29) = 1.03, p = .15. In the two-adjective condi- 2
It is likely that control subjects would also show longer
tion, she increased from 0.3 to 0.6 pauses, t(26) = onset latencies for the phrases than the sentences. As discussed
1.46, p = .08, which was only marginally sig- in the introduction, Smith and Wheeldon (1999) found that
nificant. subjects were faster to initiate single-clause sentences begin-
ning with a simple phrase than matched single-clause sentences
The types of errors for the patients’ final beginning with a complex phrase. The difference in onset for
attempts are presented in Table 4. The errors on the different sentence types was on the order of 78 ms and onset
final responses fell into four types. Most of the latencies were around 1 second for their young normal subjects.
errors consisted of producing a grammatically Although there are many differences between their experiment
appropriate phrase or sentence, but using one and ours, EA’s onset latencies and her difference between the
sentence and phrase constructions may fall close to the range of
incorrect adjective (e.g., producing ‘‘short blonde older (age-matched) control subjects. ML’s mean onset laten-
hair’’ instead of ‘‘short curly hair’’). The errors cies and differences between phrase and sentence latencies are
listed under ‘‘incorrect adjective’’ were always certain to fall far outside this range.
SEMANTIC RETENTION AND SPEECH PRODUCTION 273
TABLE 4
Percent correct and percent error types in phrase and sentence productions in Experiment 1
Controls
One adjective
Phrase 90 95 5 0 0 0
Sentence 98 98 2 0 0 0
Two adjectives
Phrase 70 81 19 0 0 0
Sentence 76 84 16 0 0 0
ML
One adjective
Phrase 78 100 0 0 0 0
Sentence 89 100 0 0 0 0
Two adjectives
Phrase 22 44 56 0 0 0
Sentence 28 34 67 0 0 0
GR
One adjective
Phrase 50 83 17 0 0 0
Sentence 83 89 6 6 0 0
Two adjectives
Phrase 28 39 56 0 6 0
Sentence 44 61 28 11 0 0
EA
One adjective
Phrase 89 100 0 0 0 0
Sentence 83 83 6 0 0 11
Two adjectives
Phrase 67 78 22 0 0 0
Sentence 89 89 11 0 0 0
onset, may reflect different strategies or degrees of phrases, they were better able to produce these
deficit for the two patients. forms. However, while their accuracy or time to
The better performance of EA than ML and onset improved, they showed greater pausing in
GR on these production tasks is consistent with the sentence forms. This greater pausing suggests
the hypothesis that the semantic-retention capa- that they produced the sentence in a piecemeal
city that has been affected in ML and GR is fashion, producing one phrase (and one content
involved in planning for speech production, pre- word) at a time and then moving on to the next.
sumably at the lexical-semantic level. Their sub- EA, with a phonological retention deficit, was
stantially better performance in the sentence than less affected by the amount of lexical-semantic
the phrase construction is consistent with the information to be maintained in a phrase, and
lexical head principle, that is, the hypothesis that showed less pausing and shorter pauses during her
all content words in a phrase up to and including sentence productions. These results suggest one of
the head of the phrase must be planned at the two possibilities with regard to the role of
lexical-semantic level prior to initiating phono- phonological retention capacity. The first is that
logical retrieval. ML and GR apparently have the phonological capacity that is affected for
great difficulty in maintaining more than one patient EA is a capacity involved in the retention
lexical-semantic representation simultaneously. of input phonological forms. The output phono-
Given that the sentence forms contained only one logical capacity that is involved in planning speech
noun in the initial noun phrase, a simple copula as production has been unaffected for this patient.
the verb, and one or two adjectives in the adjective The second is that phonological planning does not
274 MARTIN AND FREEDMAN
occur across a very long span in terms of number and binding theory of syntactic structure
of words, perhaps covering only one or two words. (Chomsky, 1981; Radford, 1988) with traces (ei)
This minimal storage demand may be within EA’s indicated for the cleft sentences:
reduced capacity. We favour the former possibi-
lity, given other evidence of a separation between (1) The rabbit chased the monkey.
input and output capacities (Martin et al., 1999; [[[the]d et [rabbit]N ]N P [[[chased]V [[the]de t
Romani, 1992). However, the second cannot be [monkey] N ]N P ]V P ] I’’ ]S
ruled out on the basis of the data presented here. (2) The monkey was chased by the rabbit.
[[[the]d et [monkey] N ]N P [[was]I [chased]V
Although the results from Experiment 1 indi-
[[by]prep [[the]det [rabbit]N ]N P ]pp ]V P ] ‘I’ ] S
cated better performance in the sentence con-
(3) That’s the rabbiti that ei chased the monkey.
structions for the patients with a semantic [[[that] N ]N P [[[is]V [[[the]det [rabbit] N ]N P
retention deficit, their errors included a large [[that]co m p [[[[chased] V [[the]det [monkey] N
proportion of incorrect adjectives in both the ]N P .]VP ]I’ ]ip ] C P’ ] N P ]V P ]IP ]S
phrase and sentence constructions, particularly for (4) That’s the monkeyi that ei was chased by the
the two-adjective condition. Patient EA also rabbit.
showed some errors of this type. This suggests that [[[that] N ]N P [[[is]V [[[the]det [monkey] N ]N P
having the comparison pictures presented along- [[that]co m p [[[was]com p [[[was]au x [[chased]V
side the target picture created interference for the [[by]prep [[the]det [rabbit]N ]N P ]PP ]PP ]V P ] I’
patients, particularly for the patients with a ]ip ] C P’ ] N P ]V P ]IP ]S
semantic retention deficit. If these patients’ diffi-
culty in production is specifically with retaining All of these sentence types allow for production in
lexical-semantic representations, then they should a phrase-by-phrase fashion in which each phrase
have difficulty producing the AN phrases even (NP, VP, PP) has at most one content word up to
when a single picture is presented. Pictures could the lexical head (i.e., the noun, the verb, or pre-
be selected with dimensions that are salient position). Even though the verb phrase in all of
enough to evoke a target response without having these examples contains two content words (e.g.,
comparison pictures. Work along these lines is ‘‘chased the monkey’’), the verb (e.g., ‘‘chased’’) is
currently in progress. the head of the phrase. According to the lexical
head principle, only the verb and words preceding
it in the same phrase have to be represented in a
EXPERIMENT 2: semantic form prior to phonological encoding. As
PRODUCTION OF SYNTACTICALLY no content words precede the verb in the same
COMPLEX SENTENCES phrase, only the verb would have to be planned.
Experiment 2 addressed the issue of whether In this experiment, actions carried out with
ML’s and GR’s difficulty with prenominal adjec- stuffed animals were used rather than pictures in
tive phrases might be attributed to difficulties with order to elicit the target productions. The
syntactically complex constructions. That is, it experimental procedure was similar to one
might be argued that producing an AAN phrase developed by Hamburger and Crain (1984) to
involves more complex syntactic planning than elicit complex constructions from children. We
producing an AN phrase, which in turn involves used this procedure because it was easier to por-
more complex planning than producing a single tray different verbs (e.g., ‘‘chased’’, ‘‘tickled’’)
noun or adjective. In this experiment, patients with this method than with static pictures.
were asked to produce grammatically complex
sentences in which the amount of semantic infor- Method
mation in each phrase was limited. The syntactic
constructions that were used were one-clause Materials
active and passive sentences (e.g., ‘‘The rabbit
Eight different stuffed animals were used to act
chased the monkey’’ or ‘‘The monkey was chased
out scenes to be described by the patient.
by the rabbit’’) and cleft-subject sentences having
either an active or passive form in the relative
Design and procedure
clause (e.g., ‘‘That’s the rabbit that chased the
monkey’’ or ‘‘That’s the monkey that was chased Before producing the sentences, subjects were
by the rabbit’’). The surface structure of the sen- familiarised with the animal names (bear, dog,
tences is shown next, according to the government alligator, fish, monkey, rabbit, frog, and elephant)
SEMANTIC RETENTION AND SPEECH PRODUCTION 275
and the actions (chase, kick, push, and tickle) that with the stuffed animals used to elicit these con-
would be used. The experimenter first named each structions. 3
animal and demonstrated each action for the
patient. Then the patient named each animal and
each action three times. Results
In the first part of the experiment, subjects
Both the AAN phrases from Experiment 1 and
produced three sets of sentences: blocked active,
these more grammatically complex sentence con-
blocked passive, and mixed active and passive. To
structions in Experiment 2 required production of
ensure that the patients understood the construc-
three content words. Table 5 compares the results
tions they were supposed to produce, they
for these tasks. The table includes results both for
received seven practice trials with feedback
initial correct responses (first correct) and those
before each blocked set, and eight practice trials
that the subject produced eventually after making
before the mixed set. There were 16 test trials in
self-corrections (last correct). Both ML and GR
each of the blocked sets, and 24 test trials in the
performed much better on the syntactically more
mixed set. The active and passive blocks always
complex sentences in the active-passive and cleft
preceded the mixed block. The experimenter
sentences than they did on the AAN phrases, both
acted out a brief scene with the animals and then
for first correct and last correct responses. Com-
pointed to one of the animals to elicit either an
paring their ‘‘last correct’’ scores, ML performed
active or a passive sentence. The subject was
significantly worse on the AAN phrases (44%)
asked to describe what occurred, beginning the
than on the mixed one-clause sentences (92%), w2
sentence with the animal pointed to by the
= 11.24, p < .001 and the mixed cleft sentences
experimenter.
(96%), w2 = 14.05, p < .001. GR also performed
For the cleft sentences, the same type of scene
significantly worse on the AAN phrases (39%)
was acted out, but there were two animals of one
than on the mixed one-clause sentences (71%), w2
kind and one animal of another kind present. One
= 4.23, p = .04. As noted earlier, he refused to
of the animals of the pair acted on or was acted
perform the mixed cleft sentence condition.
upon by the other animal. The experimenter
However, it is apparent that he did quite well on
pointed to that animal, and the patients were
the blocked active and passive cleft sentences in
instructed to describe what had occurred, begin-
comparison to the AAN phrases. Patient EA
ning each sentence with ‘‘That’s the . . .’’ in order
showed no significant difference in performance
to invoke a relative clause construction. (For
between the AAN phrases (78%), the mixed one-
example, if there were two frogs and one bear, and
clause sentences (92%) and the mixed cleft sen-
one of the frogs kicked the bear, the experimenter
tences (79%), p > .05 for all comparisons.
would point to that frog, and the target response
Table 6 reports the number of pauses and pause
would be ‘‘That’s the frog that kicked the bear’’).
lengths for these sentence types for first correct
Again, subjects received practice trials with feed-
responses. As in the sentence constructions in
back before each block, and the subject relative
Experiment 1, patients ML and GR had more
clause block and the object relative clause block
pauses, and their pauses (particularly ML’s) were
preceded the mixed block. Alternative cleft pro-
longer than EA’s in all conditions. Collapsing
ductions that were appropriate for the picture
across all correct responses in all six sentence
(e.g., ‘‘It was the frog that kicked the bear’’) were
conditions, ML made significantly more pauses
counted as correct.
than EA, t(166) = 5.54, p < .0005, and his pauses
The entire session was tape-recorded and the
were also significantly longer than hers, t(141) =
patients’ responses were digitised. Their responses
3.22, p = .002. Comparing the five sentence con-
were scored for accuracy and were timed to
dition blocks that GR completed, GR also made
measure the number and length of pauses within
significantly more pauses than EA, t(131) = 3.86, p
their responses as in Experiment 1. It was not
< .0005, and his pauses were significantly longer
possible to measure time to onset of their
than EA’s, t(100) = 3.26, p = .002.
responses because the scenes were acted out and
Although ML’s and GR’s performance on
subjects could begin planning and initiating their
these sentence constructions was better than their
response during the scene. There was no audible
signal to begin timing for each trial.
GR did not complete the mixed cleft sentence 3
It is not clear why he refused to complete this condition.
condition because he refused to do any more tests Perhaps he found the use of the stuffed animals demeaning.
276 MARTIN AND FREEDMAN
TABLE 5
Percent correct on tasks requiring production of three content words
ML
First correct 22 75 94 67 88 88 54
Last correct 44 81 100 92 88 100 96
GR
First correct 28 81 81 58 100 81 —
Last correct 39 87 81 71 100 94 —
EA
First correct 67 88 94 92 88 81 79
Last correct 78 94 94 92 88 81 79
TABLE 6
Number of pauses and mean pause length (in seconds) for Experiment 2
ML
Mean # pauses 1.1 1.3 1.1 1.3 0.9 1.8
Mean pause length 3.1 3.5 2.5 2.2 2.6 2.3
GR
Mean # pauses 1.2 0.5 1.4 1.0 1.0 —
Mean pause length 2.1 1.9 2.9 2.4 2.0 —
EA
Mean # pauses 0.5 1.1 0.4 0.0 0.5 0.5
Mean pause length 1.7 1.6 1.7 0.0 1.3 1.5
performance on the AAN phrases, they both did those responses in which the patient produced a
show some breakdown on the mixed conditions, sentence using the wrong voice.
particularly in terms of first correct responses. Although few of ML and GR’s errors on final
Table 7 repeats the first correct and last correct utterances were the wrong voice, they produced
percentages and provides a breakdown of errors many such errors in their initial attempts in the
for final utterances (i.e., for sentences that were mixed conditions. Interestingly, these initial voice
never produced correctly). These errors were errors were always passive responses to active
categorised into four types. ‘‘Wrong verb’’ refers targets, rather than active responses to passive
to trials in which the subject used a different verb targets (see Table 8). The inappropriate passive
from the one acted out in the scene (e.g., ‘‘The dog constructions usually occurred after a passive
chased the fish’’ instead of ‘‘The dog tickled the sentence had been elicited in the previous trial.
fish’’). ‘‘Verb form’’ refers to trials in which the For these incorrect passive responses, the patients
patient produced a grammatically incorrect form (especially ML) sometimes realised the form was
of the verb (e.g., ‘‘The rabbit tickle the dog’’). This incorrect and self-corrected to the correct form.
was the primary error type for ML and GR. For example, for the target ‘‘That’s the rabbit that
‘‘Sentence form’’ errors were those in which the chased the dog’’, ML responded ‘‘It was the uh,
target was a complex (cleft) sentence form but the rabbit that was chased the dog. It was the rabbit
patient produced a simple form (e.g., ‘‘The dog that chased the dog’’. (As in this example, some of
chased the fish’’ instead of ‘‘That’s the dog that these initial utterances were incomplete passive
chased the fish’’). Only EA produced this type of forms.) Patient EA did not produce a voice error
error. Finally, ‘‘Active/Passive’’ errors include on any of the trials.
SEMANTIC RETENTION AND SPEECH PRODUCTION 277
TABLE 7
Percent correct and percent error types on final utterances in Experiment 2
ML
Actives (n = 16) 75 81 0 19 0 0
Passives (n = 16) 94 100 0 0 0 0
Mixed (n = 24) 67 92 0 4 0 4
Subject Relative (n = 8) 88 88 0 12 0 0
Object Relative (n = 16) 88 100 0 0 0 0
Mixed (n = 24) 54 96 0 0 0 4
GR
Actives (n = 16) 81 87 0 13 0 0
Passives (n = 16) 81 81 0 19 0 0
Mixed (n = 24) 58 71 13 8 0 8
Subject Relative (n = 8) 100 100 0 0 0 0
Object Relative (n = 16) 81 94 6 0 0 0
Mixed (n = 24) — — — — — —
EA
Actives (n = 16) 94 94 0 6 0 0
Passives (n = 16) 94 94 6 0 0 0
Mixed (n = 24) 92 92 8 0 0 0
Subject Relative (n = 8) 88 88 0 0 12 0
Object Relative (n = 16) 81 81 13 0 6 0
Mixed (n = 24) 79 79 4 0 17 0
TABLE 8
Number of active/passive voice errors on initial utterances in Experiment 2
ML
Mixed Act/Pass 8 0 3 2 2
Mixed Sub/Obj Rel. 11 0 5 4 4
GR
Mixed Act/Pass 10 0 4 3 1
EA
Mixed Act/Pass 2 0 0 0 0
Mixed Sub/Obj Rel. 5 0 0 0 0
Although ML and GR showed a good ability to complex structures would suggest that their
produce the complex sentences in Experiment 2, capacity restrictions are not at the level of main-
their performance did drop from the blocked to taining complex syntactic frames. Their difficulty
the mixed conditions, at least in terms of first seems instead to be in maintaining the lexical-
correct responses. (Their final responses in the semantic representations that would be attached
mixed condition were highly accurate.) As dis- to such frames. Thus, these results support the
cussed in the results section, a large proportion of hypothesis that the lexical-semantic retention
their initial errors in the mixed condition were the buffer is separate from the syntactic planning
inappropriate production of a passive form for an frame, and serves to hold simultaneously several
active target. This finding is surprising given that representations that will be attached to the syn-
active forms are much more frequent in naturally tactic frame.
occurring speech and one might have predicted
the production of actives for passives rather than
the reverse. Most of these passive response errors GENERAL DISCUSSION
(75%) occurred following the correct production
of a passive form on the immediately preceding As discussed in the introduction, patients with a
trial, thus indicating a tendency to perseverate on semantic retention deficit have difficulties com-
a passive form. Studies with normal subjects have prehending sentences containing noun phrases
demonstrated a syntactic priming effect—that is, with several adjectives preceding a noun, but do
use of a given syntactic construction in one better when the adjectives follow the noun (Mar-
utterance primes the production of that same tin, 1995; Martin & Romani, 1994). In Experiment
syntactic form on the next utterance (Bock, 1986). 1, patients with a semantic retention deficit had
However, the amount of priming appears to be difficulty with the same prenominal adjective
equivalent for active and passive primes (Bock, constructions in speech production. Although
1986; Bock, Loebell, & Morey, 1992). It would they could name single nouns and adjectives
seem that it is something about the difficulty of the accurately, their performance declined sub-
passive form for these aphasic patients that leads stantially when they had to produce AN or AAN
to its persistence. It is unclear, however, why a phrases. However, their performance improved in
difficult passive form should show more persis- Experiment 1 when they could produce the nouns
tence than an active form, which should be gen- and adjectives in a sentence form with the adjec-
erally more available. tives following the noun. We have argued that the
In the introduction we discussed the possibility sentence form placed fewer demands on semantic
that the lextical-semantic retention buffer might retention because of the lexical head principle—
be equivalent to the syntactic planning frame that is, the number of content words up to and
hypothesised by Dell (1986). The results from including the head of the phrase was smaller in the
Experiment 2 argue against this possibility. That sentence condition. Experiment 2 demonstrated
is, in order for patients to produce, for example, that these patients’ difficulty with the prenominal
the cleft sentences, which contain an embedded adjective phrases could not be attributed to syn-
clause, we would assume that they must maintain tactic complexity of the phrases, as they were
something like the syntactic tree structure in the better able to produce syntactically complex sen-
sentence types shown in examples (3)–(4). Even tences including sentences with a passive form in a
though we are assuming that syntactic planning is relative clause than they could prenominal adjec-
incremental rather than carried out simulta- tive phrases with an equal number of content
neously for an entire clause, the syntactic choices words. The findings from both experiments sup-
that are made for each fragment that is planned port the contention that the same lexical-semantic
must be retained in order to ensure that the retention capacity is involved in comprehension
structure of subsequent fragments can be inte- and production.
grated with the structure of the earlier fragments To provide a comparison with the patients with
to construct a syntactically coherent sentence (de a semantic retention deficit, patient EA, with a
Smedt, 1996; Kempen & Hoenkamp, 1987). Thus, phonological retention deficit, was also included
there is a need for a capacity for retaining the in these experiments. Despite her smaller memory
syntactic structure generated so far so that sub- span, she performed better than the patients with
sequent fragments may be integrated with this a semantic retention dificit did on the AN phrases.
structure. The patients’ success at producing the She showed no significant improvement in the
SEMANTIC RETENTION AND SPEECH PRODUCTION 279
Hamburger, H., & Crain, S. (1984). Acquisition of Martin, R.C., & Romani, C. (1995). Remembering
cognitive compiling. Cognition, 17(2), 85–136. stories but not words. In R. Campbell & M. Conway
Hanten, G., & Martin, R.C. (2000). Contributions of (Eds.), Broken memories (pp.267–284). Cambridge,
phonological and semantic short-term memory to UK: Blackwell.
sentence processing: Evidence from two cases of Martin, R.C., Shelton, J., & Yaffee, L.S. (1994). Lan-
closed head injury in children. Journal of Memory guage processing and working memory: Neu-
and Language, 43, 335–361. ropsychological evidence for separate phonological
Hulme, C., Maughan, S., & Brown, G. (1991). Memory and semantic capacities. Journal of Memory and
for familiar and unfamiliar words: Evidence for a Language, 33, 83–111.
long-term memory contribution to short-term span. Meyer, A.S. (1996). Lexical access in phrase and sen-
Journal of Memory and Language, 30, 685–701. tence production: Results from picture–word inter-
Kempen, G., & Hoenkamp, E. (1987). An incremental ference experiments. Journal of Memory and
procedural grammar for sentence formulation. Language, 35, 477–496.
Cognitive Science, 11, 201–258. Patterson, K., Graham, N., & Hodges, J.R. (1994). The
Kempen, G., & Huijbers, P. (1983). The lexicalization impact of semantic memory loss on phonological
process in sentence production and naming: Indirect representations. Journal of Cognitive Neuroscience,
election of words. Cognition, 14, 185–209. 6, 57–69.
Knott, R., Patterson, K., & Hodges, J. (1997). Lexical Poirier, M., & St. Aubin, J. (1995). Memory for related
and semantic binding effects in short-term memory: and unrelated words: Further evidence on the
Evidence from semantic dementia. Cognitive influence of semantic factors in immediate serial
Neuropsychology, 14, 1165–1216. recall. Quarterly Journal of Experimental Psy-
LaPointe, S.G., & Dell, G. (1989). A synthesis of some chology, 48(A), 384–404.
recent work in sentence production. In G.N. Carlson Radford, A. (1988). Transformational grammar: A first
& M.K. Tanenhaus (Eds.), Linguistic structure in course. New York: Cambridge University Press.
language processing (pp.107–156). Boston, MA: Romani, C. (1992). Are there distinct input and output
Kluwer Academic Publishers. buffers? Evidence from an aphasic patient with an
Levelt, W., Roelofs, A., & Meyer, A. (1999). A theory impaired output buffer. Language & Cognitive Pro-
of lexical access in speech production. Behavioral & cesses, 7(2), 131–162.
Brain Sciences, 22, 1–75. Romani, C., & Martin, R. (1999). A deficit in the short-
Martin, N., & Saffran, E.M. (1992). A computational term retention of lexical-semantic information:
account of deep dyslexia: Evidence from a single Forgetting words but remembering a story. Journal
case study. Brain and Language, 43, 240–274. of Experimental Psychology: General, 128(1), 56–77.
Martin, N., & Saffran, E.M. (1997). Language and Saffran, E.M., & Martin, N. (1990). Neuropsychological
auditory-verbal short-term memory impairments: evidence for lexical involvement in short-term
Evidence for common underlying processes. Cogni- memory. In G. Vallar & T. Shallice (Eds.), Neuro-
tive Neuropsychology, 14(5), 641–682. psychological impairments of short-term memory
Martin, R. (1995). A multiple capacities view of working (pp.428–447). London: Cambridge University Press.
memory in language. Symposium presentation, Saffran, E.M., Schwartz, M.F., Linebarger, M., &
Psychonomic Society meeting, Los Angeles. Bochetto, P. (1989). The Philadelphia Comprehen-
Martin, R., & Freedman, M. (2001). Verbal working sion Battery [unpublished].
memory: The ins and outs of phonological and Schriefers, H. (1993). Syntactic processes in the pro-
lexical-semantic retention. In H.L. Roediger, J.S. duction of noun phrases. Journal of Experimental
Nairne, I. Neath, & A.M. Suprenant (Eds.), The Psychology: Learning, Memory and Cognition, 19,
nature of remembering: Essays in honor of Robert G. 841–850.
Crowder (pp.331–349). Washington, DC: American Schweickert, R. (1993). Degradation and redintegration
Psychological Association Press. in immediate recall: A multinomial processing tree
Martin, R.C., & Feher, E. (1990). The consequences of model. Memory and Cognition, 21, 168–175.
reduced memory span for the comprehension of Schweickert, R., Guentert, L., & Hersberger, L. (1990).
semantic versus syntactic information. Brain and Phonological similarity, pronunciation rate, and
Language , 38, 1–20. memory span. Psychological Science, 1(1), 74–77.
Martin, R.C., & Lesch, M.F. (1996). Associations and Shallice, T., & Butterworth, B. (1977). Short-term
dissociations between language impairment and list memory impairment and spontaneous speech.
recall: Implications for models of STM. In S.E. Neuropsychologia, 15, 729–735.
Gathercole (Ed.), Models of short-term memory Shallice, T., Rumiati, R.I., & Zadini, A. (2000). The
(pp.149–178). Hove, UK: Psychology Press. selective impairment of the phonological output
Martin, R.C., Lesch, M.F., & Bartha, M.C. (1999). buffer. Cognitive Neuropsychology, 17, 517–546.
Independence of input and output phonology in Smith, M., & Wheeldon, L. (1999). High level proces-
word processing and short-term memory. Journal of sing scope in spoken sentence production. Cognition,
Memory and Language, 40, 1–27. 73, 205–246.
Martin, R.C., & Romani, C. (1994). Verbal working
memory and sentence comprehension: A multiple-
components view. Neuropsychology, 8(4), 506–523.