See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/340741732

A Replication of Waugh and Norman (1965) Primary Memory study

Article  in  The Quantitative Methods for Psychology · April 2020

DOI: 10.20982/tqmp.16.2.r001

CITATIONS READS
2 7,924

29 authors, including:

Marilou Poitras Lucie Péléja

University of Ottawa University of Ottawa
29 PUBLICATIONS   13 CITATIONS    3 PUBLICATIONS   2 CITATIONS   

SEE PROFILE SEE PROFILE

Gardy Lavertu Anouck Langlois

Université de Montréal University of Ottawa
5 PUBLICATIONS   3 CITATIONS    2 PUBLICATIONS   2 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Pubertal lipopolysaccharide treatment disrupts hypothalamic-pituitary-gonadal-axis View project

Myocardial Infarction in women: risk factors, symptoms of MI, neuropsychological, physiological, and emotional impact View project

All content following this page was uploaded by Marilou Poitras on 18 April 2020.

The user has requested enhancement of the downloaded file.

 ¦ 2020 Vol. 16 no. 2

A Replication of Waugh and Norman (1965) Primary

Memory study

Marilou Poitras a , Lucie Péléja a , Gardy Lavertu a , Anouck Langlois a , Katia Boulerice a ,
Pauline Berthelot a B , Philippe Vincent-Lamarre b , Sandy Beaulieu a , Vanessa Bournival a , Lau-
rence Brault a , Jenna Charlebois a , Eve Camille Galloway a , Ariane Gauthier a , Rose-Marie Gibeau a ,
Nicolas Giroux a , Grace Jacob a , Myriam La Flèche a , Lély-Rose Laurin a , Véronique Legault a , Marie
Lessard a , Chelsea McAlpine a , Camille Mercier a , Albino Nikolla a , Dominique Perras a , Bryanna
Marie Pidgeon a , Sydney Running a , Kim Thériault a , Marissa Hailey Trudel a & Mikaela Rose
Winder a
a
Université d’Ottawa
b
Université du Québec à Montréal

Abstract Waugh & Norman’s experiment (1965) is among the most influential studies in the field Acting Editor De-
of cognitive psychology. Using a probe digit memory test, they proposed that the proportion of cor- nis Cousineau (Uni-
versité d’Ottawa)
rectly remembered items, digits in this case, depends on the number of interfering items shown
between the recall item and the signal, or probe, identifying the recall target. This indicates that Reviewers
interference alone accounts for forgetting in short-term memory. The following study aimed to No authors of
the origal article
replicate these results with greater statistical power as the original study used a small sample of 4 reviewed this article
participants. In a second study, we used shorter lists to examine potential effects of the relative dif-
ficulty of the task. Both studies’ results partially support Waugh & Norman’s claim, as participants
were more likely to recall a digit that was followed by fewer interfering items. Additionally, we ob-
served an interaction involving presentation rate and interference, as participants performed best
with a low amount of interfering items presented at 1 item per second. However, lists with a higher
number of interfering items (7 and above) had similar correct recall proportion regardless of the
presentation rate. These findings further support the prevalence of interference theory over the
decay theory but also call for a closer look at the possible interaction between the two. Future stud-
ies should examine the latter as well as the possible effect of cognitive fatigue due to the difficulty
of the task, underlining the importance of replication studies.
Keywords Short-term memory, Interference theory, Decay theory, Replication study.

B pbert033@uottawa.ca

10.20982/tqmp.16.2.r001

what causes forgetting in short-term memory, the de-

Introduction
cay and interference theories being the most prominent
There are various types of memory systems, such as the (Jonides et al., 2008). Waugh and Norman (1965) argued
long term, the short term and the perceptual (echoic and that interference alone causes forgetting in short-term
iconic) memories. In this article, we focus on the short- memory. It stipulates that the proportion of correctly re-
term memory, which is characterized by its very limited membered items depends on the number of interfering
capacity as well as its ability to temporarily maintain cog- items shown between the to-be-remembered target and
nitive information accessible (Cowan, 2008). Earlier stud- the cue indicating how to identify the target (a probe).
ies have shown that the short-term memory has a capacity As shown in Figure 1, left panel, the interference theory
of 7 plus or minus 2 items (Miller, 1956). However, recent predicts that the relevant factor is the number of interfer-
studies revised this capacity to roughly 4 (Cowan, 2001). ing items. Thus, varying the rate of presentation –the main
Many hypotheses have been considered to determine manipulation in Waugh and Norman (1965)– should make

The Quantitative Methods for Psychology

2 r1
 ¦ 2020 Vol. 16 no. 2

Figure 1 Visual representation of the results expected based on the theory of interference (left) and the theory of decay
(right). The curves were generated using a power curve, 0.98 × n−0.85 , where n is the number of interfering items, 0.98,
the initial performance, and 0.85, the forgetting rate (Cousineau, Hélie, & Lefebvre, 2003).

(a) (b)

no difference. Therefore, we should expect two identical list of stimuli to examine the potential effect of the diffi-
curves regardless of the rate of presentation of items. culty of the probe digit task.
On the other hand, the right panel of Figure 1 shows
Study 1
the curve expected according to the decay theory, which
states that the sole factor of forgetting is the amount of time The goal of Study 1 is to validate the mechanisms under-
elapsed. Contingent to this theory, varying the rate of pre- lying short-term memory forgetting using lists of unre-
sentation predicts different curves. However, the recall is hearsed digits, by modifying the presentation rate and the
identical for a given amount of time elapsed between the number of interfering items. Following the original study
target and the probe. (Waugh & Norman, 1965), this paradigm tests the theory of
This well-known study has been very influential and decay against the theory of interference. The procedure is
is found in nearly all textbooks on cognitive psychology. an exact replication of the classic study done by Waugh and
However, it has a major representativity limitation as it Norman (1965). The replication study was pre-registered
used a sample size of only 4 participants, constituted exclu- prior to beginning the data collection and the stimuli as
sively of Harvard undergraduate students. The APA (Van- well as the raw data for the experiment have been made
denBos, 2010) recommends that there are always at least available online (https://osf.io/hgfqy/).
20 participants per group. It is meant to minimize the risk
Methodology
that a homogeneous –but unrepresentative– sample biases
the results under cluster sampling (Cousineau & Lauren- The experimental session was divided in 4 blocks, with
celle, 2015). breaks in between each block of a duration determined by
The purpose of this replication is to verify the results the participant. Each block contained 27 trials. One trial is
obtained by Waugh and Norman (1965) because their par- composed of a list of 16 randomly generated single digits,
ticipants do not constitute a statistically representative in which the same digit could not appear more than twice
sample due to its lack of size and diversity (Faber & Fon- in a row. The lists were presented at a rate of either one
seca, 2014). Two studies were conducted. Study 1 is a direct digit per second (slow list) or four digits per second (fast
replication of the original whereas Study 2 used a shorter list). Each participant did both lists, the first 2 blocks were

The Quantitative Methods for Psychology

2 r2
 ¦ 2020 Vol. 16 no. 2

Figure 2 Results obtained in the present replication (full lines) and in the original Waugh and Norman (1965) study
(dashed line). The proportion of correct answers is presented as a function of the number of interfering items; error bars
shows correlation and difference-adjusted 95% confidence intervals of the mean (Cousineau, 2017). There are no error
bars on the original study because the raw data are not available.

presented at one randomly determined presentation rate their response.

and the following 2 blocks at the other rate. Prior to the Instructions were given verbally to the participants by
presentation of each list, a white fixation plus sign would the experimenter and were also displayed on the com-
appear on the black screen for 4 seconds, after which it puter screen prior to beginning the task. The instruc-
would disappear and the list would begin, presented audi- tions and the audio stream were presented through ear-
torily. phones, in either French or English according to the par-
The objective for the participants was to identify the ticipant’s primary language. The digits were read aloud by
number having immediately followed a probe digit, which a monotone male voice synthesized using Text2speech
could be presented only once before in the list, in posi- (https://www.text2speech.org/). The recordings were mod-
tions 3, 5, 7, 9, 10, 11, 12, 13 or 14. The probe was the last ified using the software Audacity to ensure that the sound
digit presented and was followed by a 250 ms tone to in- fit within 250 ms. The digit zero was excluded from the
dicate the end of the list. Each probe position was tested stimuli in both languages and the digit seven was excluded
6 times within each presentation rate. The participants in- from the English stimuli because two-syllable digits would
dicated the digit having immediately followed the probe not fit within 250 ms.
in the list by pressing the correct number key on the key- Participants were recruited directly by the experi-
board, guessing if unsure. Since the probe had only been menters, who were students enrolled in the University of
presented once before in the list, there was only one possi- Ottawa’s PSY3777 course of autumn 2019. There were no
ble correct answer. Nine practice lists were presented be- eligibility criteria other than for participants to not have
fore the first and the third blocks to allow participants to any memory impairment. The participants were asked
accommodate to the task. During these practice trials, the to bring their own headphones for the experiment. The
participants could see if their answer was correct; these software used for the experiment was E-Prime. The ex-
trials were not included in the analyses. During the actual perience took place in silence to avoid any distractions.
trials, participants did not receive any feedback regarding Each participant in the study was entered into a draw for a

The Quantitative Methods for Psychology

2 r3
 ¦ 2020 Vol. 16 no. 2

chance to win one of three 25$ Amazon gift cards. proportion of correct answers almost equal to 1.0 for both
presentation speeds. For the same amount of interfering
Results
items, our replication found a mean accuracy around 0.5
The participants were 24 individuals from the greater Ot- for the slower rate and around 0.3 for the faster one. The
tawa region, Ontario. The sample was composed of 63.8% small sample size of the original study could partially ac-
females and most of the participants were undergraduate count for these differences, as the original sample might
students at the University of Ottawa. The average age of have been less representative.
the participants was 21.4 years (SD = 2.2). The study was
Study 2
conducted in a computer room with 30 stations at the Uni-
versity of Ottawa. Waugh and Norman derived their experiment by taking
The mean proportion of correct recall is given in Fig- into account a model of short term memory with two pro-
ure 2 as a function of the number of interfering items, cesses in which repetition serves both to maintain the
separated for fast presentation rate and slow presentation items in primary memory and eventually to record them
rates. For comparison, we also added (dashed lines) the in a secondary memory; the modal model (Watkins, 1974).
results of the original study. However, the original methodology with the initial lists
Informal examination indicates no effect of the lan- of 16 digits does not take into consideration the memory
guage in which the stimuli were presented and likewise no span of 7 plus or minus 2 items of the short-term memory
effects of gender. (Miller, 1956).
A one-way within-subjects analysis of variance We therefore considered that it would be interesting to
(ANOVA) comparing the effect of number of interfering bring some modifications to Waugh and Norman’s (1965)
items on short-term memory recall in both slow and fast original study. The goal was to observe whether the re-
presentation rate found a significant effect of presentation sults would remain the same with shorter lists and ascer-
rate, F (1, 22) = 32.90, p < 0.001. Moreover, the effect of tain that the performances in the original study were not
interfering items on memory recall also yielded a signifi- partly caused by the presentation of too many digits which
cant effect of interference, F (8, 176) = 7.84, p < 0.001. would have made the task too difficult.
Finally, there was also a significant interaction between In the original study, there was an important effect of
rate of presentation and number of interfering items, the number of interfering items, yet no effects were re-
F (8, 176) = 2.47, p = 0.015. ported regarding the rate of presentation (1 item/sec and
4 items/sec). However, a study by Altmann and Schunn
Discussion
(2012), showed that the two factors interact more than
The results obtained from this experiment partially sup- what was initially believed. Indeed, when the rate of pre-
port Waugh and Norman’s (1965) conclusion that only in- sentation is fast, the first items in the list benefit from the
terference affects retention in short-term memory. It can reduced effect of decay. Conversely, when presentation
be noted that the average proportion of correct recall de- is slow, the later items benefit from the decay of previ-
creases with the number of interfering items, as predicted ous items and the resulting reduction in proactive inter-
by Figure 2. However, the curves are distinct at first and ference.
then converge to an asymptotic performance at about 20% To examine this point, we tested these two presenta-
correct recall. Whereas the decay theory predicts that the tion rates on shorter lists to observe if the influence would
two curves would diverge, we see the inverse effect. be the same on the modal model illustrated by Waugh and
When there were less interfering items, participants Norman as with the original lists of 16 digits.
had overall a better performance when the listed numbers The theories of decay and interference were contrasted
were presented at a rate of one digit per second than a rate by using two different presentation rates. Also, having less
of 4 digits per second. However, as the number of inter- interfering items in this study aimed to test the theory of
fering items increased, performance for both presentation interference and how much it would influence the partic-
rates tend to be similar. Although this supports the inter- ipants’ results. This study is not an exact replication of
ference theory, the fact that at a slower rate performances Waugh and Norman’s (1965) original study as the lists in-
are higher indicates another possible variable causing this cluded fewer items.
distinction.
Methodology
A notable difference between our results and those ob-
tained in the original experiment is the starting mean ac- Study 2 aimed at evaluating once again the mechanisms
curacy. As presented in Figure 2, at a minimum of three of short-term memory, this time using lists of eight unre-
interfering items, Waugh and Norman (1965) reported a hearsed digits, instead of sixteen. Each block included 25

The Quantitative Methods for Psychology

2 r4
 ¦ 2020 Vol. 16 no. 2

Figure 3 Mean proportion recalled across participants as a function of the number of interfering items for both presen-
tation rates in Study 2 (full lines) and the original experiment (dashed lines).

trials of eight randomly generated single digits, with the for the slower presentation rate is superior to the faster
constraint that the same digit could not appear more than presentation rate. A weaker performance is noted when
twice in a row. The probe was the 8th digit presented, fol- the number of interfering items increased (3 to 6).
lowed by a 250 ms tone to indicate the end of the list. A two-way repeated measure ANOVA revealed a sig-
As in Study 1, the objective of the task was to identify nificant main effect of presentation rates on proportion
the number having immediately followed the probe digit, of correct recall (F (1, 29) = 25.85, p < 0.001). Par-
which was presented only once before in the list, in posi- ticipants showed an increased performance in the 1 item
tions 2, 3, 4, 5 or 6. The position of the probe digit and the per second compared to the 4 items per second condi-
length of the lists are the only differences in methodology tion. It also revealed a significant main effect of the num-
between the two studies. Participants were entered in the ber of interfering items on proportion of correct recall
same draw as those of Study 1 for a chance to win one of (F (4, 116) = 13.12, p < 0.001). Participants’ performance
three 25$ Amazon gift cards. For further details, refer to improved when there were less interfering items follow-
the methodology section of Study 1. ing the probe digit. Finally, there was a significant interac-
tion between presentation rate and number of interfering
Results
items (F (4, 116) = 4.01, p = 0.004) such that participants
Study 2 had a total of 29 participants of which 69.1% (N = scored the highest when they were in the 1 item per second
20) were women. Most subjects were undergraduate stu- condition with a small number of interfering items follow-
dents from University of Ottawa and their age ranged from ing the probe digit. However, this effect seems to disappear
17 to 25 years old (M = 20.56, SD = 1.9). Most partic- when the probe digit is at the very beginning of the series
ipants completed the test in a University of Ottawa com- (interfering items = 7). Participants scored similarly with 7
puter laboratory with 30 stations; 5 did it alone in an indi- interfering items regardless of the presentation rate.
vidual booth.
Discussion
Figure 3 shows the proportion of correctly recalled
items as a function of the number of interfering items. When comparing our results with those obtained by
When there are fewer interfering items, the performance Waugh and Norman (1965), we noticed that for the larger

The Quantitative Methods for Psychology

2 r5
 ¦ 2020 Vol. 16 no. 2

number of interfering items (6), there seems to be no dif- number of participants as well as the limited amount of
ference between the two. However, a faster presentation is time given to replicate the original study, multiple partic-
associated with a reduced recall performance for a smaller ipants were tested in one location at the same time. This
number of interfering items. This is contrary to the results decreased the amount of control the experimenters had on
that would be predicted by the decay theory. The impli- the experimental setting and increased the amount of bi-
cations and suspected causes of this phenomenon will be ases. For instance, the amount of distractions both in the
further discussed in the next section. It is also possible surroundings, such as movement or conversations, and in
to graphically observe that the data of the last interfer- the participants’ own vicinity, such as their phone, could
ing item of both presentation rates are almost layered on have decreased concentration on the task. Over 30 exper-
top of each other which could potentially be the result of imenters also took part in the making of this study which
a primary effect. The participants’ average performance increases the variability of the experimental flow from one
in study 2 is lower than the average obtained by Waugh & participant to the other. Furthermore, because some par-
Norman but higher than the one obtained in study 1. ticipants were given shorter lists than others, they finished
the task sooner and were free to leave. This could have ex-
General discussion
erted a social pressure on the remaining participants with
In Study 1, performance declines, reaching a floor, regard- longer lists who were unaware of the time difference be-
less of presentation speed, after 6 interfering items. Av- tween the two experimental conditions. It is possible that
erage performance thereafter is just slightly above chance this additional pressure further decreased performance.
level, indicating that accurate recall was highly unlikely.
Authors’ note
This coincides with the maximal capacity of short term
memory of 7 plus or minus 2 items reported in the liter- The first six authors have contributed equally to the article.
ature (Miller, 1956). With the results found in study 2, we
References
can see that performance for both presentation speeds also
rejoins after 5 or 6 interfering items. This, in combination Altmann, E. M., & Schunn, C. D. (2012). Decay versus in-
with the higher average recall when the length of the list is terference: A new look at an old interaction. Psy-
reduced such as in study 2, accentuates the concerns that chological Science, 23(11), 1435–1437. doi:10 . 1177 /
presenting more than 8 digits surpasses the capacities of 0956797612446027
human short term memory. Cousineau, D. (2017). Varieties of confidence intervals. Ad-
Although this replication addressed the main limita- vances in Cognitive Psychology, 13, 140–155. doi:10 .
tion of the original study by having a statistically more 5709/acp-0214-z
valid sample size, other limitations still apply. The sample Cousineau, D., Hélie, S., & Lefebvre, C. (2003). Testing cur-
size, albeit much larger than the original, was still mainly vatures of learning function on individual trial and
composed of young adults pursuing post-secondary educa- block average data. Behavior Research Methods, In-
tion. Additionally, while the original sample was composed struments, & Computers, 35, 493–503. doi:10 . 3758 /
mainly of men, our replication sample was mainly com- bf03195528
posed of women. These factors may impact the represen- Cousineau, D., & Laurencelle, L. (2015). A correction factor
tativeness of our sample and affect the generalizability of for the impact of cluster randomized sampling and
our results. its applications. Psychological Methods, 21, 121–135.
As shown by the reduced performance in the replica- doi:10.1037/met0000055
tion, it is possible that the task was too difficult. Due to Cowan, N. (2001). The magical number 4 in short-term
the challenge of presenting 4 digits per second, it is possi- memory: A reconsideration of mental storage ca-
ble that the quality of the audio was affected, which nega- pacity. Behavioral and Brain Sciences, 24(1), 87–114.
tively impacted performance. This could partially account doi:10.1017/S0140525X01003922
for the observed differences between the fast and slow Cowan, N. (2008). What are the differences between long-
presentation rates, as the slower presentation rate might term, short-term, and working memory? Progress in
have been clearer. Due to the difficulty of the task, par- brain research, 169, 323–338. doi:10 . 1016 / S0079 -
ticipants might also have been prone to becoming discour- 6123(07)00020-9
aged, which could have reduced the amount of effort made Faber, J., & Fonseca, L. M. (2014). How sample size influ-
to correctly answer each list. ences research outcomes. Dental press journal of or-
The experimental setting, which presents notable dis- thodontics, 19(4), 27–29. doi:10.1590/2176- 9451.19.4.
tinctions from the original study, might also account for 027-029.ebo
the deviation from the expected results. Due to the higher

The Quantitative Methods for Psychology

2 r6
 ¦ 2020 Vol. 16 no. 2

Jonides, J., Lewis, R. L., Nee, D. E., Lustig, C. A., Berman, VandenBos, G. R. ( (2010). Publication manual of the Amer-
M. G., & Moore, K. S. (2008). The mind and brain ican psychological association (6th ed.) Washington,
of short-term memory. Annual review of psychology, DC: American Psychological Association.
59(1), 193–224. doi:10.1146/annurev.psych.59.103006. Watkins, M. J. (1974). Concept and measurement of pri-
093615 mary memory. Psychological Bulletin, 81(10), 695–711.
Miller, G. A. (1956). The magical number seven plus or Retrieved from https://doi.org/10.1037/h0036952
minus two: Some limits on our capacity for process- Waugh, N. C., & Norman, D. A. (1965). Primary mem-
ing information. Psychological Review, 63, 81–97. Re- ory. Psychological Review, 72, 89–104. doi:10 . 1037 /
trieved from https://doi.org/10.1037/h0043158 h0021797

Open practices

The Open Material badge was earned because supplementary material(s) are available on https://osf.io/hgfqy/.
The Preregistered badge was earned because the experiment(s) was preregistered at https://osf.io/hgfqy/).

Citation

Poitras, M., Péléja, L., Lavertu, G., Langlois, A., Boulerice, K., Berthelot, P., . . . Winder, M. R. (2020). A replication of Waugh
and Norman (1965) Primary Memory study. The Quantitative Methods for Psychology, 16(2), r1–r7. doi:10 . 20982 /
tqmp.16.2.r001

Copyright © 2020, Poitras et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The
use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication
in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with
these terms.

Received: 27/03/2020 ∼ Accepted: 13/04/2020

View publication stats

The Quantitative Methods for Psychology
2 r7