Mechanisms of Verbal Short-Term Memory

CURRENT DtRECTIONS

iN PSYCHOtCKIICAL SCIENCE

problem it poses for traditional chaining theory, in

Quanl/tat/ve Analyses of Behavior: Discrimination

Processes,

M,L. Commons, R.|. Herrnstein, and

A,R. Wagner, Eds, (Ballinger, Cambridge, MA,

1984),

10.

H,S, Terrace, Chunking during serial learn-

ing by a pigeon: I, Basic evidence, journal of

Exper-

imental

Psychology:

Animal Behavior

Processes,

17,

61-93 (1991); M.R. D'Amato and M, Colombo,

Representation of serial order in monkeys {Cebus

apella),

journal of

Experimental

Psychology: Animal

Behavior Processes, U, 131-139 (1988); K,B,

Swartz, S. Chen, and H,S, Terrace, Serial learning

by rhesus monkeys: I, Acquisition and retention of

multiple four-item lists, journal of

Experimental

Psy-

chology: Animal Behavior

Processes,

17, 396-410

(1991).

11.

B.O. McConigle and D. Neapolitan, 5er/a/

learning in children, manuscript submitted for pub-

lication (1994),

12.

M,R. D'Amatn and M, Colombo, Serial

learning wUh wild card items by monkeys (Cebus

apella).

Implications for knowledge of ordinal

posi-

tion,

journal of Comparative Psychology, 15, 252—

261 (1989),

13.

S, Chen, K, Swartz, and H.S, Terrace,

Knowledge of ordinal position ol list items in

rhesus

monkeys, manuscript submitted for publication

(1994).

14.

S, Chen, K, Swartz, and H,S, Terrace,

Serial

learning in the monkey II: The

emergence

trial

and error

strategy

during the mastery of

successive

lists, manuscript submitted for publication (1991),

15.

P,E, Bryant and T. Trabasso, Transitive in-

ferences and memory in young children. Nature,

232. 456-^58 (1971),

16.

B,O, McConigle and M, Chalmers, Are

monkeys logical? Nature, 267, 355-^377 (1977);

6,O, McConigle and M. Chalmers, Monkeys are

rational!

Quarterly lournal of

Experimental

Psychol-

ogy, 45B, 189-228(1992),

17.

M. Harris and B.O. McConigle, Modelling

transitive inference. Quarter// lournal of

Experimen-

tal

Psychology,

47B, 319-348 (1994).

18.

B.O. McConigle and M. Chalmers, The se-

lective impact of question form and input mode on

the symbolic distance effect in children, journal of

Experimental

Child

Psychology,

37, 525-554

(1984),

19.

B, Inhelder and

Piaget, The Early Growth

Logic

in the Child (Routledge & Kegan Paul, Lon-

don,

1964),

20,

M, Chalmers and B.O. McConigle, An ex-

perimental analysis of size seriation skills in

chil-

dren, manuscript submitted for publication (1994).

21,

B,O, McConigle and M, Chalmers, An Ex-

perimental Analysis of Ordering Skills in Children,

Economic and Social Science Research Council

Project Grant Report (British Library, London,

1993),

22,

B.O, McConigle and C. De Lillo,

The

spon-

taneous regulation of

strategies

in a spatial search

task

by Cebus apella, manuscript submitted for pub-

lication (1994).

23,

B, St. lohnston, from transitive inference to

exhaustive seared; Towards self-regulatory models

of developmental

processes,

unpublished doctoral

dissertation, Edinburgh University, Edinburgh

(1994),

24,

S, Sternberg, S, Monsell, and C.E. Wright,

The latency and duration of rapid movement se-

quences: Comparisons of speech and typewriting, in

information

Processing

in Motor Control and learn-

ing,Ct. Stefmach, Ed, (Academic Press, New York,

1978),

Mechanisms of Verbal Short-Term

Memory

Nelson Cowan

A fundamental concept in cogni-

tion is that only a small amount of

information can be kept in mind at a

particular moment. William James'

labeled this phenomenon "primary

memory," which is now commonly

called "short-term memory." These

terms have been used in contrast to

"secondary" or "long-term" mem-

ory, which refer to the much larger

amount of information saved in the

brain for later retrieval. Most psy-

chologists consider short-term mem-

ory to be important in much of cog-

nition,

playing a key role in the

"working memory" system that

holds onto the information needed

Nelson Cowan is Professor of Psy-

chology at the University of Mis-

souri,

Columbia. Address corre-

spondence to Nelson Cowan,

Department of Psychology, 210

McAlester

Hall,

University of Mis-

souri,

Colunnbia, MO

65211;

e-mail: psycowan@mizzou1 .mis

souri.edu.

for comprehension and problem

solving.^

Yet the empirical support for

short-term memory as a distinct

mechanism has been weak enough

that many psychologists have

doubted the concept. Across many

years,

researchers have debated

whether memory tests over short and

long periods of time tap into differ-

ent processes, or whether instead

the results of all memory experi-

ments can be described by a unitary

set of rules that is valid regardless of

the delay imposed between the stim-

ulus presentation and memory test.^

According to the unitary memory

hypothesis, for example, remember-

ing a phone number for a few sec-

onds is accomplished by the same

rules of learning as memorizing and

later reciting the Pledge of Allegiance.

Though the passage of time

would cause forgetting according to

either hypothesis, the proposed

mechanisms of forgetting differ.

Both hypotheses allow that the pas-

sage of time can remove the subject

from the context in which the infor-

mation to be recalled was encoun-

tered originally. However, the uni-

tary memory theorists rely on that

sort of principle exclusively to ex-

plain forgetting, whereas the short-

term memory theorists assume in ad-

dition that a transient representation

of some aspects of stimulation, such

as the detailed sensory and phonetic

properties of spoken words, exists

but is lost within several (up to about

30) seconds, unless the transient

representation is renewed before

that time through some sort of re-

hearsal.

This review discusses recent

research that supports the short-term

memory hypothesis in the verbal do-

main and describes relatively spe-

cific ways in which short-term mem-

ory might operate.

FINDINGS FROM SPAN AND

SUPRASPAN

RESEARCH TRADITIONS

Most of the evidence for verbal

short-term memory storage has

come from two basic types of proce-

dures:

span and supraspan. In mem-

ory span procedures, lists of unre-

lated verbal items are to be recalled

immediately after their presentation.

The lists increase in length until sub-

VOlUMh

\^,

NUMBER 6, DECEMBER 1994

jects no longer can recall them, and

the longest lists correctly recalled

serve as approximate indicators of

short-term memory capacity. In su-

praspan procedures, the subject is

also asked to recall lists of unrelated

verbal items, but more items are pre-

sented on each trial than the subject

can recall. Researchers have found

that the most recent items on the list

are more likely to be recalled than

items earlier on the list (the

recer)cy

effect).

Some investigators have

taken this phenomenon to reflect a

contribution of short-term memory,

though this specific conclusion has

been questioned rightfully, as I ex-

plain later.

Verbal memory span experiments

have revealed an intriguing, brief

time limit in recall. Specifically,

given a particular subject and stim-

ulus set, the mean span is roughly

equal to the largest amount that the

subject can pronounce, speaking

quickly, in about 2 s. This relation

has been obtained across variations

in the subject's age, in the language,

and in the length of words in the list

to be recalled."*

Alan Baddeley explained this lin-

ear relation between speech rate and

merpory span by positing an articu-

latory-loop mechanism in short-term

recall.^

According to this account, a

phonological representation of each

list item is stored in short-term mem-

ory automatically, but fades within 2

unless it is refreshed through covert

verbal rehearsal. It is assumed that

the faster one can pronounce the

items,

the faster one can rehearse

them covertly and, therefore, the

more items one can rehearse in a

repeating loop without losing the

short-term representation. This type

of account depends inevitably on

the assumption that there is a distinct

mechanism for short-term (transient)

memory storage.

Unlike memory span procedures,

supraspan procedures allow investi-

gations of the pattern of errors across

serial positions. For 10 or more

years,

it was thought that the

privi-

leged recall of the most recently pre-

sented items in a list indicated a con-

tribution of a separate short-term

memory mechanism. A finding

taken to support that account was

that the recency effect was elimi-

nated when a period of distracting

activity (lasting, e.g., 20 s) separated

the list and the recall periods. The

usual interpretation was that the in-

terposed activity allowed the short-

term representation of the list items

to decay.^

This view of supraspan experi-

ments has been disproved by studies

using a "separated-item" testing

procedure. In that procedure, there

is a period of distracting activity not

only after the last list item, but after

each list item. Despite the final dis-

tracting period, which should dis-

rupt short-term storage, the recency

effect is obtained with this proce-

dure.''

Accounts of these results typ-

ically favor a unitary mechanism

that is common to immediate and

separated-item recall, such as the

enhanced temporal distinctiveness

of the final items in a list.^

There is much left to reconcile

between the two views. The unitary

memory theorists have not provided

an alternative account of the linear

relation between speech rate and

memory span that could operate

without a transient storage mecha-

nism.

The short-term memory theo-

rists,

for their part, have not proven

that a particular model of span tasks

is correct and have not demon-

strated that short-term storage is

needed to explain the results of su-

praspan tasks. However, recent re-

search with both span and su-

praspan tasks shows that short-term

memory exists, and reveals some of

its mechanisms.

SHORT-TERM MEMORY LOSS

DURING VERBAL

OUTPUT DELAYS

In the conventional form of the

articulatory-loop theory, the exis-

tence of a short-term memory store

can only be inferred indirectly. A

correlation between articulatory rate

and memory span is accounted for

with a proposed causal relation be-

tween covert rehearsal speed and

memory span, and rapid covert re-

hearsal is assumed to be helpful be-

cause it counteracts short-term

memory loss. However, recent

stud-

ies suggest that there may be more

direct ways of observing short-term

memory loss and its role in process-

ing during spoken recall.

Though short-term memory may

be lost while the subject is silently

receiving the list to be recalled, be-

cause of a limit in covert rehearsal

speed,^

my colleagues and 1^ dem-

onstrated that short-term memory is

lost also while the subject is recall-

ing the list aloud. While the first

word is being repeated, the other

words cannot be rehearsed and may

be forgotten; while the second word

is being repeated, the remaining

words may be forgotten; and so on.

Our demonstration was a detailed

analysis of the word-length effect,

the well-known finding that words in

a list can be recalled more success-

fully when they are phonologicaliy

shorter and therefore can be pro-

nounced more quickly."^ We hy-

pothesized that short words are ad-

vantageous for recall not just

because they can be rehearsed more

quickly than longer words, but also

because short words do not impose

as much of a delay in overt verbal

recall.

In our study, which used a su-

praspan procedure, each printed list

could begin with either short or long

words and, independently, could

end with either short or long words.

At the end of each list presentation,

there was a cue signaling a forward

or a backward order of recall. To the

extent that the word-length effect oc-

curs because of forgetting during

verbal responding, the length of

whatever words are to be recalled

first should be more important than

the length of words to be recalled

Published by Cambridge University Press

later (because the words recalled

earlier delay the recall of a larger

number of additional words than do

those recalled later).

The results, shown in Figure 1,

were as predicted. For the recall of

the entire list, the word lengths that

mattered most (producing significant

effects within a recall order) were

the lengths of the words that were to

be recalled early on

(i.e.,

the first

words in the list for forward recall

and the last words in the list for

backward recall).^ The data suggest

that word length modulated memory

loss largely during overt recall, and

not only during input and covert re-

hearsal.

During overt spoken recall,

moreover, memory loss can be ex-

amined in some detail.

DIRECT SUPPORT FOR THE

SHORT-TERM

g MEMORY CONCEPT

The word-length effect, unlike

many of the phenomena that tradi-

tionally had been taken to reflect a

contribution of short-term memory,

never had been examined using a

separated-item technique, so that

was our next

step.

Our version of the

short-term memory hypothesis at-

tributes the word-length effect

largely to greater short-term memory

loss during spoken recall when

longer words are pronounced. Ac-

cording to our hypothesis, the word-

length effect should not occur if

there is a filled delay at the end of

the list, such as the one imposed in

the separated-item procedure, be-

cause the delay should result in the

dissipation of all inform.ation in

short-term storage, leaving no short-

term memory to decay subsequently

during spoken recall. The word-

length effect should not occur with

this procedure because performance

should reflect only the retrieval of

information saved in long-term

memory. In contrast to this hypoth-

esis,

unitary memory theorists could

Fig.

1. Proportion correct as a function

of the length of words in the first two

(left-hand panels) and the last two (right-

hand panels) of five serial positions

(S.P.).

Results are shown separately for

forward recall (top panels) and back-

ward recall (bottom panels). The length

of the third (middle) word varied ran-

domly within each condition, and there

were filler lists to prevent subjects from

detecting a pattern. Length effects were

significant only for

the

length of

words

whichever half of the list was to be re-

called first (top left

and

bottom right pan-

els).

Data from Cowan et al.^

expect word-length effects in sepa-

rated-item recall similar to those ob-

tained in immediate recall, given

that these theorists have placed great

importance on the fact that results

for separated-item recall so far have

resembled the results of immediate

recall.

We tested these predictions re-

cently in an experiment'" using

backward recall and manipulating

word length separately in the first

and second halves of the list. With a

separated-item procedure, the ad-

vantage for short words was not ob-

tained,

whereas with an immediate-

recall procedure, it was. These

results reconfirm that there is a pro-

cess,

presumably short-term storage,

that can be used only in immediate-

memory tasks. The word-length ef-

fect depends on such a short-term

process rather than on unitary

prin-

ciples that apply to any kind of

memory test regardless of the time

periods involved.

MECHANISMS OF

SHORT-TERM MEMORY LOSS

AND REACTIVATION

In the supraspan task, the clues to

underlying mental processes are

simply the failures in recall. Using

the span task, however, it is possible

to gain further clues to mental pro-

cesses from the timing of correctly

recalled lists. Recently, my col-

leagues and I have taken this less-

traveled route." We set out to dis-

tinguish between two models of

short-term memory loss during the

recall period In a span task, shown

in Figure 2. In the first model, mem-

ory is lost steadily throughout the re-

call period, as several researchers

have

assumed.

In the second model,

memory loss occurs only while the

subject is pronouncing an item in re-

call.

Then, during interword pauses

in the recall period, something hap-

pens to refresh the short-term mem-

ory representations of some items.

The evidence to date suggests that

the second model is correct

and,

fur-

ther, that the covert process during

interword pauses in recall may be a

rapid memory search undertaken to

determine which item is to be pro-

nounced next.

Model 1: Steady Loss

Model 2: Loss & Reactivation

Time In Spoken Recall Pence

Fig.

2. Two possible models of short-

term memory loss during the spoken re-

call period in a memory span task:

steady memory loss (top panel) and loss

during words, with reactivation during

pauses (bottom panel). After Cowan.'^

According to the steady-loss

model,

the basis of individual differ-

ences in memory span presumably is

that some subjects can speak more

quickly during recall than others,

and thus can recall more items dur-

ing the fixed, limited period of short-

term memory availability. Our work

with children ages 4 through 8 ap-

pears to contradict this account,

however.^' Correct spoken recall

was not at all limited to a fixed, 2-s

period.

Instead, subjects with differ-

ent memory spans recalled their

span-length lists at comparable rates

of about 1 s for each correctly re-

called item, resulting in longer last-

ing recall for more capable subjects.

Older children can speak more

quickly than younger ones, but they

do not do so in the recall task. Thus,

these data lend support to the loss-

and-reactivation model of short-term

memory.

Further clues to the processing

taking place during the recall period

came from examining the duration

of word responses and interword

pauses in recall.'^ First, in the sub-

jects tested (children ages 4 through

8),

though the number of syllables in

each word in the list affected the du-

rations of words spoken in recall, the

number of syllables did not affect the

durations of subjects' interword

pauses in recall. Given the seem-

ingly ubiquitous finding of word-

length effects whenever verbal

rehearsal is used, the absence of

word-length effects on interword

pauses suggests that the covert pro-

cesses carried out in the pauses gen-

erally do not include rehearsal, at

least not in the age group studied.

Instead,

a more rapid memory

search process may be carried out.

Previous studies of memory search

focusing on adults' reaction times in

reporting whether a probe item ap-

peared in a short list also failed to

obtain word-length effects.'^

Second,

in our studies, a subject's

memory span was found to bear no

relation to the duration of words in

the subject's spoken recall," in con-

trast to the clear effect of the length

of stimulus words on word durations

in the responses. However, subjects

with higher memory spans had

shorter interword pauses in the re-

call of lists of a particular length.

Moreover, for a given subject, the

pauses were shorter within shorter

lists than within longer lists. To illus-

trate these relations. Figure 3 shows

some data from the two studies in

which interword pauses in verbal re-

call were examined." Data Bars A

and B are from my study of individ-

ual differences in recall for monosyl-

labic words in 4-year-olds. Inter-

word pauses within the responses to

three-word lists were longer in sub-

jects for whom this was the span

length (A) than in more advanced

subjects for whom this was one

word below span (B). Data Bars C

and D are from a developmental

study, averaged here across lists of

monosyllabic, bisyllabic, and trisyl-

labic words. The pauses were longer

for 4-year-olds recalling their span-

length lists (C) than for 8-year-olds

Data

A. 3-word lists In 4-year-oldE with a span of 3 words

B. 3-word lists in 4'year'Olds with a span ol 4 words

C. Span-length lists in 4-year-olds (mean list langtti: 3(

D. Lists I item tielow span in B-year-olds (mean list

Fig.

3. Effects of individual differences,

age,

and list length on interword pauses

in spoken recall. For a particular list

length,

subjects who had a lower span

and

C) made longer interword pauses

than subjects who had a higher span (B

and D). The X marks above Data Bars B

and D indicate the mean lengths of in-

terword pauses for those same subjects

in span-length lists.

The

variables shown

in this figure did not affect the durations

of words during recall, whereas the

number of syllables in each word af-

fected word durations but not interword

pauses. Data Bars A and B from

Cowan;"

Data Bars C and D from

Cowan et al.," collapsed across lists of

short, medium, and long words.

recalling lists that were of a similar

mean length but one word below

their span (D). For data bars reflect-

ing below-span list lengths (B and

D),

the Xs above the bars show how

long interword pauses were for the

same subjects in span-length lists.

Clearly, though interword pauses in-

creased with list length, for a partic-

ular list length, the pauses were

shorter in more capable or older

children.

Such effects on interword

silent intervals might be attributed to

a speedup in covert processes that is

known to occur

children mature.^^

In sum, whereas the original ar-

ticulatory-loop model assumed that

word-length and age effects on

memory span both work by modu-

lating the speed of covert rehearsal,

the output timing research indicates

that the mechanisms of age and

word-length effects cannot be

iden-

tical.

Word length affects the dura-

tion of words in the recall response,

whereas age affects instead the du-

ration of interword pauses.

We are not yet sure what pro-

cesses take place during interword

pauses. However, the literature pro-

vides clues. Measures of the timing

of verbal responses in experiments

conducted by Saul Sternberg and his

associates'"^ resemble the studies I

have discussed so far, but with adult

subjects only, subspan lists only,

and the requirement to speak as

quickly as possible. A large number

and variety of

such

experiments pro-

duced results suggesting to Sternberg

that while responding, subjects con-

duct a rapid memory search be-

tween words to determine which

word to say next, but that the search

includes all items in the list, even

items that have been recalled al-

ready. For example, as the number

of items in the list increased, inter-

word periods became longer, but

the durations ofthe internal portions

of words did not change. (In Stern-

berg's study, the interword period

for the word pair "copper, token,"

for example, would be defined as

the period including the "-er" of

Published by Cambridge University Press

CURRENT

OmCTlONS iN PSYCHOLOGICAL SCIENCE

"copper" and the "t-" of "token,"

and the internal portions would be

the remaining parts ofthe response.)

In addition, interword durations

were constant across serial

posi-

tions.

In all of these respects, the re-

sults appear to agree with our own

results in the span task.

The processes that affect memory

during the spoken recall period in

memory span tasks, in children at

least, are consistent with the bottom

panel of Figure 2 and a working

model that might be summarized as

follows. The length of stimulus

words affects how much time

passes, and therefore how much of

short-term memory is lost, while

words are spoken during recall. The

subject's mnemonic capability helps

to determine how quickly or

effi-

ciently memory search can occur

during silent pauses in recall, and

this memory search may, as a useful

byproduct, reactivate some items in

short-term memory; faster reactiva-

tion in more capable or older sub-

jects would prolong short-term

memory activation in these subjects,

and thus can explain why their span-

length recall responses last much

longer than those of less capable or

younger subjects.

I must caution that the effects de-

scribed here are likely to be based

on only some of the mental pro-

cesses available to subjects to carry

out short-term memory tasks. The ef-

fects occur most reliably when the

stimuli on each trial are drawn from

a small set and recall in the correct

serial order is required. Even under

such circumstances, though, there

are some few subjects who may be

able to circumvent the short-term

memory mechanisms I have dis-

cussed by using special strategies,

such as rapidly forming semantic as-

sociations to reorganize the materi-

als to be recalled.'^

CONCLUSIONS

There is a set of phenomena spe-

cific to short-term memory proce-

dures that cannot be explained by

unitary rules that have been pro-

posed to apply to all memory tasks.

These phenomena include the ef-

fects of word length and age: Longer

words allow more loss of informa-

tion from short-term memory not just

by slowing the rate of covert re-

hearsal,

but also by slowing the rate

of overt spoken recall. Older sub-

jects are able not just to rehearse

more,

but also to search through

memory more efficiently, decreas-

ing the length of pauses in recall and

possibly increasing the rate of re-

freshment of items in a short-term

store.

The unitary memory concept

cannot explain these findings easily

because they depend on the specific

amount of time in which stimulus in-

formation can be lost from short-

term memory, and not only on more

general factors such as the temporal

distinctiveness of stimuli.

Given that there is a fundamental

human limitation in short-term

memory and in skills that rely on it,

it is important to know what causes

the limitation. Recent research dem-

onstrates that, although a unitary set

of rules describing learning is impor-

tant, a special, short-lived memory

representation also plays an impor-

tant role.

Acknowledgments—This work was sup-

ported by Grant HD-21 338 from the Na-

tional Institutes of Health. I thank Judith

Goodman,

Robert Logie, Noelle Wood,

and two anonymous reviewers for com-

ments on an earlier draft.

Notes

W. Jafne5, The Principles of Psychology

(Henry Holt, New York, 1890).

). Cantor, R.W. Engle, and G. Hamilton,

Short-term memory, working memory, and verbal

abilities: How do they relate? Intelligence, 15, 229-

246 (1991): R.H. Logie, K.J. Gilhooly, and V.

Wynn,

Counting on working memory in arithmetic

problem solving, Memory & Cognition (in press).

3. For the two-process view, see N. Cowan, Ac-

tivation,

attention, and shor!-lerm memorv. Memory

& Cognition, 2!, 162-167 (1993). For the unitary

view, see R.G. Crowder, Short-term memory:

Where do we stand? Memory & Cognition, 21, 142-

145 (1993).

C. Hulme and V. Tordoff, Working memory

development: The effects of speech rate, word

length,

and acoustic similarity on serial recali,

lour-

nai of Experimental Child Psychology, 47, 72-87

(1989), and references therein.

5. A.D. Baddeley, Working Memory (Clarendon

Press, Oxford, 1986).

6. M. Glanzer and A.R. Cunitz, Two storage

mechanisms in free recall, journal of Verbal Learn-

ing and Verbal Behavior, 5, 351-360 (1966).

7. R.A. Bjork and W.B. Whitten, Recency-

sensitive retrieval processes in long-term free recall,

Cogn/I/ve Psychology, 6, 173-189 (1974); R.L.

Greene, Sources of recency effects in free recall.

Psychological Bulletin, 99. 221-228 (1986). Greene

also describes certain other effects that previously

had been attributed to the contribution of a short-

term storage mechanism.

8. R.G. Crowder, Principles of Learning and

Memory (Erlbaum, Hillsdale, N), 1976); A.M. Glen-

berg and N.C. Swanson, A temporal distinctiveness

theory of recency and modality effects, journal of

Experimental Psychology: Learning, Memory, and

Cognition, 12, 3-15 (1986).

9. N. Cowan, L. Day, J.S. Saults, T.A. Keller, T.

Johnson,

and L. Flores, The role of verbal output

time in the effects of word length on immediate

memory, joumal of Memory and Language, 31,

1-17 (1992); for a convergent result, see L.A.

Henry, The effects of word length and phonemic

similarity in young children's short-term memory.

Quarterly joumal of

Experimental

Psychology, 43A,

35-52 (1991).

10.

N, Cowan, N.L, Wood, and D.N. Borne,

Reconfirmation of the short-term storage concept,

Psyc/io/ogical Science, 5, 103-106 (1994).

11.

N. Cowan, Verbal memory span and the

timing of spoken recall, lotimal of Memory and

Lan-

guage,

31, 668-684 (1992); N. Cowan, T. Keller, C.

Hulme, S. Roodenrvs, S. McDougall, and |. Rack,

Verbal memory span in children: Speech timing

clues to the mechanisms underlying age and word

length effects, journal of Memory and

Language,

33,

234-250(19941.

12.

W.G. Chase, Does memory scanning in-

volve implicit speech? in Attention and

Performance

VI, S. Dornic, Ed. (Erlbaum, Hillsdale, NJ, 1977),

and references therein.

13.

S. Hale, A global developmental trend in

cognitive processing

speed.

Child Development, 6J,

653-663 (1990), and references therein.

14.

S. Sternberg, C.E, Wright, R.L. Knoll, and S,

Monsell,

Motor programs in rapid speech; Addi-

tional evidence, in Perception and Production of

Fluent

Speech,

R.A. Cole, Ed. (Ertbaum, Hillsdale,

NJ,

1980), and references therein.

15-

S. Delta Sala, R.H. Logie, C. Marchetti, and

V. Wynn, Case studies in working memory: A case

for single cases? Cortex, 27, 169-191 (1991),