ArticlePDF Available

Is the Fast-Same Phenomenon That Fast? An Investigation of Identity Priming in the Same-Different Task

May 2022
Journal of Experimental Psychology Learning Memory and Cognition 48(4)

May 2022
48(4)

DOI:10.1037/xlm0001076

Authors:

Brad Harding

Université de Moncton

Denis Cousineau

University of Ottawa

The same-different task is a classic paradigm that requires participants to judge whether two successively presented stimuli are the same or different. While this task is simple, with results that have been replicated many times, response times (RTs) and accuracy for both same and different decisions remain difficult to model. The biggest obstacle in modeling the task lies within its effect referred to as the fast-same phenomenon whereby participants are much faster at responding "same" than "different," while most standard cognitive models predict the opposite. In this study, we investigated whether this effect is the result of identity priming activated by the first stimulus. We ran four variants of the same-different task in which identity priming is intended to be attenuated or cancelled in half of the trials. Results for all four variants show that a complete visual match between both stimuli is necessary to observe a fast-same effect and that hampering this relation attenuates same RTs while different RTs remained relatively unchanged. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

…

Figures - uploaded by Brad Harding

Content may be subject to copyright.

Content uploaded by Brad Harding

Content may be subject to copyright.

Is the Fast-Same Phenomenon That Fast? An Investigation of Identity

Priming in the Same-Different Task

Bradley Harding

and Denis Cousineau

École de Psychologie, Faculté des sciences de la santé et des services communautaires, Université de Moncton

École de Psychologie, Université d’Ottawa

The same-different task is a classic paradigm that requires participants to judge whether two successively pre-

sented stimuli are the same or different. While this task is simple, with results that have been replicated many

times, response times (RTs) and accuracy for both same and different decisions remain difﬁcult to model. The

biggest obstacle in modeling the task lies within its effect referred to as the fast-same phenomenon whereby

participants are much faster at responding “same”than “different,”while most standard cognitive models pre-

dict the opposite. In this study, we investigated whether this effect is the result of identity priming activated by

the ﬁrst stimulus. We ran four variants of the same-different task in which identity priming is intended to be

attenuated or cancelled in half of the trials. Results for all four variants show that a complete visual match

between both stimuli is necessary to observe a fast-same effect and that hampering this relation attenuates

same RTs while different RTs remained relatively unchanged.

Keywords: same-different task, matching task, comparison task, priming, fast-same phenomenon

At any given moment, we are faced with an incredible number

of stimuli, forcing many decisions. Sorting out stimuli that remain

unchanged from one moment to the next is an efﬁcient way to

minimize the number of operations. While it is known that com-

parison processes are performed very efﬁciently, little is known

regarding how “sameness”between stimuli is detected with split-

second latency and near-perfect accuracy (see Farell, 1985, and

Sternberg, 1998, for extensive reviews). To explore the question,

the same-different task (sometimes called the comparison task or

the matching task), a simple task where participants must judge as

accurately and as rapidly as possible whether two presented stim-

uli are the same or different, can be used.

The Same-Different Task

There are many variants of the same-different task, all with the

goal of assessing how we compare pairs of stimuli, including

variants in which there is a comparison of letters (e.g., Bamber,

1969,1972;Bamber & Paine, 1973;Krueger, 1973;Nickerson,

1965;Taylor, 1976a), numbers (e.g., Silverman & Goldberg, 1975;

Snodgrass, 1972;Van Opstal & Verguts, 2011), words (e.g., Farell,

1977;Well et al., 1975), faces (e.g., Megreya & Burton, 2006;Tver-

sky, 1969), abstract patterns (e.g., Bindra et al., 1968;Dyer, 1973;

Egeth, 1966;Hock, 1973;Link & Tindall, 1971;Nickerson, 1967a,

1967b;Nickerson & Pew, 1973;Snodgrass, 1972;Taylor, 1969),

motion direction (Petrov, 2009), and tones (Bindra et al., 1968;

Bindra et al., 1965). There also exists two variants that affect the

task’s decision rule: (a) In the conjunctive, or “all-same”task (Bam-

ber, 1969;Derks, 1972), participants answer “same”when the crite-

rion stimulus (S

)andtheteststimulus(S

) match on all attributes;

“different”responses are invited when at least one attribute differs.

(b) In the less-common disjunctive, or “all-different”task, partici-

pants answer “same”as soon as a single match between S

and S

found; for this task, participants must answer “different”if, and only

if, all attributes mismatch (Nickerson, 1967a;Sekuler & Abrams,

1968;Silverman & Goldberg, 1975;Taylor, 1976a;aswellasFarell,

1977, reviewed thoroughly in Farell, 1985).

Herein, all same-different tasks will have a conjunctive deci-

sion rule, and the compared stimuli will be strings of letters

sampled from the Latin alphabet; these stimuli will be presented

in quick succession. Thus, each experiment aims to replicate

Bamber’s (1969) seminal experiment that sparked decades of

research and debates that revolved around the task’s most nota-

ble and robust result, the fast-same phenomenon. Henceforth, we

will describe stimulus composition using the dDlL nomenclature

in which drepresents the number of mismatching letters (D,

the “differing”letters) between the S

-S

pair and lrepresents

the total number of letters (L) composing both stimuli. For exam-

ple, a 1D3L condition represents a pair of letter strings in which

there is a single difference between the three letters composing

This article was published Online First May 12, 2022.

Bradley Harding https://orcid.org/0000-0003-1222-8769

Denis Cousineau https://orcid.org/0000-0001-5908-0402

This research has been funded by the New Brunswick Innovation

Foundation and a fellowship from the National Science Research and

Engineering Council of Canada attributed to Bradley Harding as well as a

research grant attributed to Denis Cousineau from the same organization.

We thank Tom Busey, Marc-André Goulet, David Huber, and Randall

Jamieson for their helpful comments on a previous version of this article.

Correspondence concerning this article should be addressed to Bradley

Harding, École de Psychologie, Faculté des sciences de la santé et des

services communautaires, Université de Moncton, Campus Moncton,

Pavillon Léopold-Taillon 18 Avenue Antonine-Maillet, Moncton, NB E1A

3E6, Canada. Email: Bradley.Harding@uMoncton.ca

520

Journal of Experimental Psychology:

Learning, Memory, and Cognition

ISSN: 0278-7393 https://doi.org/10.1037/xlm0001076

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

and the three letters composing S

. The location of the differ-

ence is random and uncontrolled. As another example, a 0D4L

condition represents a pair of strings in which there are no differ-

ences between the four letters composing S

and the four letters

composing S

(a “same”condition). It is not possible to have a

condition where dexceeds las there cannot be more differences

than there are letters. Hence, the design is not fully crossed. In

these experiments, there are four possible “same”conditions

(0D1L to 0D4L) and 10 possible “different”conditions (1D1L to

4D4L). For simplicity’s sake, we will refer to these as the stimu-

lus composition conditions.

The Fast-Same Phenomenon

The fast-same phenomenon (expression ﬁrst found in Bamber,

1972, p. 321, but with allusions found in Bamber, 1969;Egeth,

1966;Nickerson, 1967a,1967b,1968) is the observation that

overall “same”response times (RTs) are reliably faster than the

overall RTs for “different.”In particular, the “same”RTs are

always faster than those of the slowest “different”conditions (for

reviews, see Farell, 1985;Nickerson, 1978;St. James & Eriksen,

1991;Sternberg, 1998). Quantitatively, there is no agreed-upon

standard as to what constitutes a fast-same effect as there has been

neither a formal proposition nor a quantiﬁed measure of the effect.

Yet, considering the literature, we can expect a moderate-to-large

Hedges’gof about .4 (Hedges’gbeing the unbiased variant of

Cohen’sd, the latter of which tends to overestimate effect sizes;

Goulet-Pelletier & Cousineau, 2018) between the overall mean RT

for “same”(the average of the four “same”stimulus composition

conditions) and the overall mean RT for “different”(the average

of the 10 “different”stimulus composition conditions).

This effect is counterintuitive from a modeling standpoint as

“same”responses should be based on an exhaustive examination

of all attributes, whereas “different”responses can be self-termi-

nating; this heuristic holds whether processing is performed seri-

ally or in parallel (Harding et al., 2016;Taylor, 1976a;Townsend

& Ashby, 1983;Townsend & Nozawa, 1995). This effect is fur-

ther characterized by (a) “same”RTs being as fast as (or faster

than) all “different”RTs and (b) very high accuracies for all

“same”conditions (often 95% and over), surpassing the overall ac-

curacy of most “different”conditions (this component of the fast-

same effect is sometimes referred to as the false-different effect;

Beller, 1970;Krueger, 1978).

In addition to the fast-same phenomenon, as noted in Stern-

berg’s (1998) review of the task’s main results, the slopes of each

“different”condition’s RTs (i.e., the 1DlL, 2DlL, 3DlL, and 4DlL

stimulus composition conditions) “fan out”as a function of stimu-

lus length. Within this fan, the steepest slope corresponds to the

line linking the stimulus composition conditions in which there is

only a single difference between the criterion and test stimuli, fol-

lowed by the line linking stimulus composition conditions where

there are two differences between both stimuli, followed by the

line linking stimulus composition conditions with three differen-

ces, and so on

. Conversely, the slope of the line linking the four

“same”stimulus composition conditions is commonly found to be

the ﬂattest and forms the bottom extremity to the fan-out effect

(Bamber, 1969,1972;Sternberg, 1998;Taylor, 1976a). Slope

analyses are relevant to understand the same-different task because

a self-terminating process is expected to predict half the slope of

an exhaustive process when detecting a single difference between

strings (the 2:1 prediction under serial processing; Cousineau &

Larochelle, 2004;Wolfe, 1994). Yet slopes are found to be larger,

not reduced, for “different”conditions when compared to the slope

of the “same”condition.

The fast-same phenomenon is robust to variations in experimen-

tal design and has become a staple ﬁnding of the task. Yet, to this

day, there is no agreed-upon explanatory model of the mechanism

(s) behind this puzzling phenomenon.

Modeling the Phenomenon

Two major lines of research were put forth in the 1970s and

1980s to explain the fast-same effect. In the ﬁrst series of experi-

ments, researchers modiﬁed the visual attributes between the com-

pared stimuli without changing the overall response modalities.

For example, Bamber (1972) used lowercase and uppercase conso-

nants and instructed participants that stimuli such as “a”and “A”

were to be considered as “same”(also see Bamber & Paine, 1973;

Posner & Mitchell, 1967;Proctor, 1981). The results of this task

variant showed slower than usual “same”RTs that still remained

faster than the slowest “different”condition. Such ﬁndings oppose

template matching models (Egeth, 1966) and dual process models

giving a preferential processing rate to identical stimuli (such as

the identity reporter model in Bamber, 1969). In the second series

of experiments, researchers manipulated response biases without

affecting the compared stimuli per se. This endeavor, led largely

by the work of Ratcliff and Hacker (1981), explored the integra-

tion of the speed–accuracy trade-off (ﬁrst introduced by Henmon,

1911; see Heitz, 2014, for a review) within the same-different task

by varying the levels of “cautiousness”participants had to exercise

before answering “same.”Their results showed an attenuation of

“same”RTs as well as a shortening of “different”RTs (Ratcliff &

Hacker, 1981).

From these two empirical approaches emerged a very fertile

exchange between Ratcliff and colleagues and Proctor and col-

leagues (in chronological order: Proctor, 1981;Proctor & Rao,

1982;Proctor & Rao, 1983a;Proctor et al., 1984;Proctor, 1986;

Ratcliff, 1985;Ratcliff & Hacker, 1981;Ratcliff & Hacker, 1983).

The ﬁrst team, using the diffusion model (Ratcliff, 1978)toinstanti-

ate their views, argued that the task can be explained solely by the

presence of response biases and trade-offs carrying varying amounts

of noise (expanding on the work of Krueger, 1978,1979; see also

Howell & Stockdale, 1975;Taylor, 1976b). The second team, not

denying possible response biases and trade-offs, argued that a deci-

sion facilitation process stemming from visually identical pairs of

stimuli was necessary to fully encapsulate the results. In this present

research, we thoroughly address the latter of the two propositions,

while the former has also been recently addressed (Goulet & Cous-

ineau, 2020b).

Visual Identity and Priming

The results of same-different tasks (Donderi & Zelnicker, 1969;

Posner & Boies, 1971) have often been contrasted with results stem-

ming from stimuli repetition tasks (Bertelson, 1961;Kornblum, 1969).

One study found that the trend continues to at least seven attributes for

simultaneous presentation (Derks, 1972).

RESIDUAL ACTIVATION PRIMING 521

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Proctor (1981) used a condition in which participants had to answer

“same”or “different”when comparing visually matching stimuli and

another condition in which the stimuli were visually mismatching

(much like Bamber’s, 1972, research). He found that both conditions

presented faster responses for repeated stimuli, but the beneﬁtwas

more pronounced for the visually matching trials. This facilitation

effect, sometimes assimilated to a priming effect (as envisioned by

Krueger & Shapiro, 1981;Nickerson, 1978;butseeProctor & Rao,

1983b), was explained by the fact that a visually identical stimulus had

already been encoded just moments prior, resulting in encoding facili-

tation akin to what is now referred to as residual activation (Huber,

2008;Huber & O’Reilly, 2003). This view would also explain why

the decision time for visually mismatching “same”stimuli, despite

being slower than their visually matching counterparts, remain faster

than the slowest “different”condition (the 1DlL condition; Bamber,

1972;Bamber & Paine, 1973;Posner & Mitchell, 1967). Indeed,

regardless of the letter’s visual identity, there still exists a phonological

and semantic relationship between stimuli (“j”and “J”are both pro-

nounced identically and represent the same letter identity-wise), result-

ing in higher-level priming. This ﬁnding led Proctor (1981) to posit

that priming is likely at play on the phonological and semantic levels

as well within the constructs of this task. This proposition, used to

explain fast-same results, was indirectly supported by Miller and

Bauer’s (1981) work on redundant attribute target detection tasks that

later led to the creation of the coactive model of decision-making

(Miller, 1982).

Although there are different priming methods, the most relevant

form for same-different decisions is identity priming, where it is

assumed that residual activation beneﬁts the processing of an iden-

tical stimulus when it is presented within a brief time interval

(Huber, 2008;Huber & O’Reilly, 2003;Jacob et al., 2013). Iden-

tity priming also posits the presence of a processing hierarchy

through which any visually presented stimulus must travel. The

bottom levels process visual information, the middle levels process

phonological information, and the top levels process semantic in-

formation (see Eviatar et al., 1994;Huber, 2008;Huber &

O’Reilly, 2003;Lupker et al., 2015, for work pertaining to target-

ing priming within speciﬁc levels of the processing hierarchy). For

a primed, identical stimulus, the processing network quickly reac-

tivates and “fast-tracks”the stimulus through the hierarchy. As

discussed, if one were to remove visual priming beneﬁts (by alter-

ing the visual aspect of S

), phonological and semantic priming

beneﬁts would possibly remain, resulting in higher-level forms of

response facilitation and consequently faster recognition of the tar-

get. In other words, stimulus processing would not beneﬁt from

the residual activation originating from the lower, perceptual lev-

els even though there may be priming beneﬁts at the upper—pho-

nological or semantic—levels.

Cancelling Identity Priming

As previously discussed, the priming model is a parsimonious

and elegant mechanism to explain the fast-same phenomenon.

However, to further validate this hypothesis, it is necessary to

create experiments in which fast-same responses are cancelled,

or at least attenuated by manipulating the strength of identity pri-

ming. As noted above, one way to do so is by altering the visual

appearance of the compared stimuli so that a different pathway

is taken to reach the upper processing levels of a decision. Such

manipulations will be found in Experiments 1 and 2 of this arti-

cle. Alternatively, it should be possible to create an alternative

mental representation of S

by changing the stimulus’s encoding

modality. For example, one could present the criterion stimulus

audibly so that participants can still create a mental representa-

tion of S

without beneﬁting from the perception of a visually

identical stimulus. This experimental manipulation is found in

Experiment 3. Finally, one can avoid identity priming altogether

by not presenting a criterion stimulus at all. Instead, cues can be

presented that allow the retrieval of S

from long-term memory

(LTM). This ensures that any activation resulting from the cues

in the perceptual levels of processing is completely unrelated to

the test stimulus. This variant is found in Experiment 4. Henceforth,

the experimental manipulations that control whether primed or

unprimed stimuli pairs will be presented will be referred to as the

priming conditions.

In this article, we explore these priming cancellation techniques

and their effect on same-different task results. If the fast-same phe-

nomenon is indeed caused by the repetition of identical stimuli, a

cancellation, or at least an attenuation, of the fast-same effect is

expected, whereas “different”decision times should remain rela-

tively unchanged. Furthermore, it is unclear whether the fast-same

phenomenon is an all-or-naught effect or if it is graded; this is one

unanswered question that will be examined here.

In all experiments, we analyzed RTs and accuracy rates for each

stimulus composition condition for both priming conditions.

Within these priming conditions, we also estimated slopes and

intercepts for “same”(0DlL) and 1DlL conditions; we chose these

two sets of stimulus composition conditions because, as noted by

Sternberg (1998), they are located at both extremities of the RT

fan-out effect. While at opposite ends of the fan empirically, they

are theoretically the most similar conditions processing-wise: The

four “same”stimulus composition conditions should follow an ex-

haustive stopping rule, and the four 1D stimulus composition con-

ditions should follow the closest to an exhaustive stopping rule of

all “different”stimulus composition conditions. It is expected that

the usual mean RT trends (Bamber, 1969; summarized in Stern-

berg, 1998) will be found in all experiments except for those in

which identity priming is altered—for these experimental variants,

we only expected changes in “same”RTs as “different”conditions

should not beneﬁt from residual activation at all (encoding a “dif-

ferent”S

forcibly uses a different neural pathway as there is at

least one mismatching letter between the criterion and test strings).

Furthermore, to extend the analyses beyond mean RT, we ana-

lyzed in Appendix A the mean standard deviation and mean skew-

ness of all 14 stimulus composition conditions for both priming

conditions within each of the four experiments. This analysis is

relevant to decode whether primed and unprimed stimuli pairs are

processed with qualitatively different decision-making mecha-

nisms. To further explore this notion, we also modeled the results

using the EZ diffusion model (Wagenmakers et al., 2007) and

present our parameter estimates in Appendix B.

Experiment 1: Case Manipulation

In this ﬁrst experiment, we varied the letter cases between S

and S

to see whether the change in appearance between stimuli

affected “same”RTs. It was expected that a fast-same effect will

occur when the stimuli are visually matching and that the mean

522 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

RTs for “same”will be attenuated when identity priming is

removed (where S

and S

are visually mismatching, akin to Bam-

ber, 1972). More critically, it was also expected that “different”

RTs will be unaffected by the priming manipulation.

Method

Participants

Participants were undergraduate and graduate students recruited

from the University of Ottawa. All participants were between 18 and

30 years of age, had normal or corrected vision, and were informed

of the experiment’s procedure as well as the protocol and ethical

rules of the University of Ottawa (REB Certiﬁcate H10-16-14). All

participants gave written and verbal consent to participate in this task

and were compensated $10 CAD for their time (approximately 1 hr

for brieﬁng, testing, and debrieﬁng). In this, and all subsequent

experiments, we aimed to recruit 20 participants, which is 5 times

more than in Bamber's (1969) original study. Statistical power to

detect a typical fast-same effect should be above 80% (with an aof

.05 and an effect size of .36, with considerations toward replicated

measures; Goulet & Cousineau, 2019).

Stimuli

Stimuli were displayed on a calibrated CRT monitor (1024 3

768; 85 Hz) to ensure luminance and RGB standards across par-

ticipants. Participants were seated approximately 50 cm away

from the center of the screen with the computer keyboard placed

on the desk directly in front of them. Participants could move the

latter slightly to ensure a comfortable position. Twelve conso-

nants (B, C, D, F, J, K, L, N, S, T, V, and Z; the same as Bam-

ber’s, 1969, original study) presented in the Courier New font

(e.g., JCVD; the same font was also used for Experiments 3 and 4)

were selected to serve as stimuli. The stimuli were presented

within a 10° 310° visual angle centered on the screen; the ﬁrst

string (S

) was shown 4° above the center of the computer screen,

and the second string (S

) was shown 4° below the center of the

screen. Stimuli were always presented as white letters on a black

background.

The letters were randomly selected on every trial. String length

(L) also varied from one to four letters on every trial. No letter was

presented twice within the same stimulus, and to simplify the com-

parison process, matching letters would appear in the same position

for both S

and S

.For“different”conditions, the mismatching let-

ter(s) were nonidentical from those already used in both S

and S

;

could have no differences (“same”) on half the trials or a number

of differences (D) between 1 and L on the other half.

Considering the priming conditions, in the visually matching

condition, there was no change in letter case between stimuli. In

this priming condition, there were 384 total trials, of which half

used lowercase letters for both S

and S

and half used uppercase

letters for both stimuli. Therefore, for “same”trials within this pri-

ming condition, the visual information was completely identical

between criterion and test letter sets. In the other 384 trials, the

visual mismatching condition, letter cases between stimuli mis-

matched between S

and S

so that on 192 trials, a lowercase S

was presented with an uppercase S

, and on 192 trials, an upper-

case S

was presented along with a lowercase S

. Unlike Bamber

(1972), the entire string’s composition was in uppercase or in

lowercase; Bamber's original experiment could have stimuli

resembling “JcvD,”or “jCvD,”which led to higher-than-normal

error rates (a pilot study using this manipulation proved to be difﬁ-

cult for participants, and mean accuracy rates plummeted below

what is typically observed). Finally, regarding both priming condi-

tions within this experiment, participants were speciﬁcally

instructed to ignore the stimuli’s visual appearance and base their

“same”and “different”responses solely on the letters’identity.

Procedure

During the on-screen instructions, participants were instructed

to respond by pressing the “CTRL”key located on the far-left side

of the keyboard using their left hand and the “ENTER”key

located on the far-right side of the keyboard (on the numeric pad)

using their right hand. The “same”or “different”decision associ-

ated with each key was counterbalanced based on the participants’

randomly assigned participant number. Therefore, half of all par-

ticipants pressed “same”with their left hand and half with their

right. The experiment began once the participant was ready and

verbally acknowledged their understanding of the procedure.

The timeline of a typical trial is shown in Figure 1. As shown,

there was a 505-ms blank screen before the trial began, followed by

aﬁxation cross presented for 505 ms. S

followed immediately after

ﬁxation and was presented on-screen for 400 ms. Afterward, a

blank screen was presented for 400 ms followed by S

, the test stim-

ulus. This test stimulus was shown for 5,000 ms or until a decision

was made. Feedback was only given for 505 ms for nonresponses

and errors to avoid diverting the participants’gaze when they

answered correctly; for correct answers, the screen was left blank

for 505 ms. While the task is easy, and participants rarely made

mistakes, a red warning message was shown if the participant made

ﬁve mistakes in a row (the warning message, which reintroduced

the basic instructions of the task, could only be ended by pressing

the space bar, a separate key to all other task responses). Finally,

participants were offered short breaks after every 192 trials (one

quarter of the experiment’s total number of trials). Following test-

ing, all participants were subjected to a debrieﬁng period to answer

any queries and to explain the goal of the study.

Experimental Design

One session consisted of 768 total trials. Both priming conditions

consisted of 384 trials each, of which half were “same”and half

were “different.”Strings of all lengths had an equal frequency of pre-

sentation, meaning that there was an equal number of 1L, 2L, 3L,

and 4L stimuli within the entire task. Additionally, within a string of

a given length, differences had the same frequency of occurrence.

For example, when a stimulus string was four letters long, 1D, 2D,

3D, and 4D trials each occurred an equal number of times, with the

position of the difference(s) assigned at random. All trials were pre-

sented in a random order. Table 1 summarizes the stimulus composi-

tion conditions (string length and number of differences) for a total

of 384 trials (representing one of the two priming conditions).

Results

Screening of the Data

Data from 14,592 total trials were gathered (768 trials 319 par-

ticipants). Twenty total participants were initially recruited, but

RESIDUAL ACTIVATION PRIMING 523

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

one was excluded prior to analysis for having very slow response

times (mean RT of 1,220 ms); it is suspected that the participant

did not understand the task instructions to respond as quickly as

possible because most of the RTs were well above 1,000 ms, an

abnormality in the task. For the remaining participants, all

responses were recorded within the allowed 5,000 ms. We

excluded 27 RTs below 200 ms and 17 RTs above 2,500 ms.

There were no nonanswers. For RT analyses, erroneous trials were

ﬁltered out to arrive at a total of 13,870 correct trials (678 errors

were recorded, representing a 4.6% error rate). Of the retained par-

ticipants, there were no instances in which participants made ﬁve

mistakes in a row. These screening procedures were identical for

all subsequent experiments.

Effect of Letter Case Differences

Because we were not interested in the speciﬁc letter case used

(uppercases vs. lowercases) but rather in the overall priming con-

dition of the responses (matching vs. mismatching), we performed

a preliminary 14 32 repeated-measures analysis of variance

(ANOVA) to identify whether there were any signiﬁcant differen-

ces in mean RT across letter case conditions (i.e., if there was a

difference between lowercase/lowercase and uppercase/uppercase

pairs of stimuli). The factors for this ANOVA are the stimulus

composition conditions (14 levels) and the letter case-matching

conditions (two levels). We found no signiﬁcant differences

between letter case conditions, F(1, 17) = .121, p= .732, h

.007 (after ﬁltering, there were no observations left in the 1D4L

condition for one participant, resulting in 17 degrees of freedom

rather than 18). We also found a strong effect of stimulus composi-

tion, F(13, 221) = 17.471, p,.001, h

= .507, and a

nonsigniﬁcant interaction between factors, F(13, 221) = .798, p=

.661, h

= .045. We performed another 14 32 repeated-measures

ANOVA to see if both mismatching conditions were different

from one another (uppercase/lowercase and lowercase/uppercase

pairs of stimuli). The factors for this ANOVA are the stimulus

composition conditions and the letter case-mismatching condi-

tions. We found a nonsigniﬁcant effect between the letter case-

mismatching conditions, F(1, 18) = .19, p= .892, h

= .001, a sig-

niﬁcant effect of stimulus composition, F(13, 234) = 14.206, p,

.001, h

= .441, and a nonsigniﬁcant interaction between the two

factors, F(13, 234) = .990, p= .461, h

= .052. Consequently, we

ignored the speciﬁc letter cases used for subsequent analyses; both

letter case-matching conditions were combined to form the matching

priming condition, and both letter case-mismatching conditions were

combined to form the mismatching priming condition.

Figure 1

Timeline of a Trial in Experiments 1, 2, and 4

Note. S

denotes the ﬁrst stimulus presented to participants (the criterion stimulus), and S

denotes the second

stimulus (the test stimulus). Feedback was only presented on errors and nonresponses. Within the actual experi-

ments, ER was the word “Error”and NR was the words “No response detected,”both shown in the lower part

of the display below S

. See the online article for the color version of this ﬁgure.

Table 1

Number of Trials as a Function of String Length for Both Same

and Different Trials

String length Same

Different

1D 2D 3D 4D

14848

2482424

3 48 161616

4 48 12121212

Total 192 192

Note. With “different”trials, there could be 1 difference (1D) or more

depending on the string’s length. This design is used for all experiments.

This table represents half of the total number of trials, equivalent to one of

the priming conditions.

524 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Mean Response Times and Accuracy

Mean RT and mean accuracy rates for each stimulus composi-

tion condition are presented in Figure 2 where statistics are shown

as a function of string length—individual lines denote the results

for the number of differences between S

and S

. The results for

each priming condition are shown in different columns. Error bars

denote the difference- and correlation-adjusted 95% conﬁdence

intervals of the mean (Cousineau, 2005;Morey, 2008) as recom-

mended by Baguley (2012; see Cousineau, 2017).

Regarding “different”RTs, the results show the expected trend

of stimulus composition for both the matching and mismatching

priming conditions: When the number of differences between S

and S

was held constant (the individual lines), RTs became slower

as the number of letters within the string increased; for any given

string length, decisions were faster as the number of differences

between S

and S

increased. We also see the slope fan-out effect

involving the stimulus composition conditions where the steepest

slope as a function of stimulus length expectedly belongs to the line

linking the four 1D stimulus composition conditions. The overall

mean RT for “different”responses (the mean RT of the 10 stimulus

composition conditions) is 547 ms in the matching condition,

whereas it is 558 ms in the mismatching condition, a difference of

11 ms (the 95% CI of this difference being 611.58 ms), this dif-

ference between overall means is nonsigniﬁcant, F(1,18) = 3.823,

p=.066,h

=.175.

A1032 repeated-measures ANOVA on mean “different”RTs

with a factor for the stimulus composition conditions and one for

the priming conditions showed a main effect of priming, F(1, 18) =

6.232, p= .022, h

= .257, a ﬁnding that goes against our initial hy-

pothesis that only “same”RTs would be affected by the change in

priming. However, the raw effect is rather small (11 ms). As

expected, there is a strong effect of stimulus composition, F(9,

162) = 24.194, p,.001, h

= .573, and the interaction between

factors is nonsigniﬁcant, F(9, 162) = .710, p= .699, h

= .038.

Regarding “same”mean RTs, in the matching condition, results

are typical of a fast-same effect. However, an identical conclusion

cannot be made for “same”trials in the mismatch condition. In

this condition, mean RTs for “same”are no longer the fastest

among all stimulus composition conditions. The overall mean RTs

for “same”responses (the mean of the four stimulus composition

conditions) is 502 ms in the matching condition, whereas it is 537

ms in the mismatching condition, a difference of 35 ms (with a

95% CI of 615 ms), which is more than 3 times the change

observed between priming conditions for “different”trials; this

difference between overall means is signiﬁcant), F(1, 18) =

22.962, p,.001, h

= .561.

A432 repeated-measures ANOVA on “same”mean RTs

revealed a main effect between the priming conditions, F(1, 18) =

22.950, p,.001, h

= .560, and a main effect of stimulus composi-

tion, F(3, 54) = 30.222, p,.001, h

= .627, but more notably, a

Figure 2

Mean Response Time (RT; Top Panels) and Mean Accuracy (Bottom Panels) Results for

Experiment 1

Note. In the left column are the results for the matching condition, and in the right column are the results for

the mismatching condition. Error bars denote the correlation- and difference-adjusted 95% conﬁdence interval

of the mean (Baguley, 2012). D = differences. See the online article for the color version of this ﬁgure.

RESIDUAL ACTIVATION PRIMING 525

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

signiﬁcant interaction between the two factors, F(3, 54) = 5.655, p=

.002, h

= .239. This signiﬁcant interaction highlights the fact that

the RT slope as a function of string length for mismatching “same”

trials is steeper than the RT slope for matching “same”trials.

Comparing overall “same”mean RTs to overall “different”

mean RTs, Hedges’gin the matching condition is .55 (with a 95%

CI of [.36, .80]; the difference in RT between these overall means

is signiﬁcant), F(1, 18) = 41.556, p,.001, h

= .698, whereas it

is .28 in the mismatching condition (with a 95% CI of [.05, .54];

this difference between overall mean RTs is signiﬁcant as well,

albeit with a smaller effect size), F(1, 18) = 5.167, p= .036, h

.223. This attenuation in Hedges’gindicates an overall attenuation

of the fast-same effect when identity priming is annulled yet still

indicates that “same”responses remain quicker than “different”

responses overall.

The mean accuracy rates for “different”responses are 95.2%

and 94.6% for matching and mismatching conditions, respectively;

for “same”responses, they are 96.5% and 95.1% for matching and

mismatching conditions, respectively. While these accuracy rates

are similar, we performed a 10 32 repeated-measures ANOVA

on “different”responses’accuracy using arcsine-transformed

mean accuracy rates. This ANOVA revealed no signiﬁcant differ-

ence between the priming conditions, F(1, 18) = 1.895, p= .186,

= .095, a main effect of stimulus composition, F(9, 162) =

32.176, p,.001, h

= .641, and a nonsigniﬁcant interaction

between factors, F(9, 162) = 1.525, p= .143, h

= .078. For

“same”responses, we performed a 4 32 repeated-measures

ANOVA on arcsine-transformed accuracy as a function of the four

stimulus composition conditions and both priming conditions. We

found a signiﬁcant difference between both priming conditions, F

(1, 18) = 6.461, p= .020, h

= .264, a signiﬁcant effect of stimulus

composition, F(3, 54) = 3.400, p= .024, h

= .159, and a signiﬁ-

cant interaction between factors, F(3, 54) = 4.294, p= .009,

= .193. This signiﬁcant interaction stems from the 0D4L

condition being more accurate in the matching condition than in

the mismatching condition (97.0% and 94.0% for matching-0D4L

and mismatching-0D4L conditions, respectively).

The conditions with the largest number of errors are the 1D

stimulus composition conditions, which are also the slowest.

Hence, this suggests that errors may be tolerated by participants to

avoid response times that would be too long in the most difﬁcult

experimental conditions.

RT Slopes and Intercepts

To see if the difference between priming conditions has deeper

roots, we measured the slope and intercept of “same”responses

and compared them to the slope and intercept of 1DlL responses

(to use as a reference) for both matching and mismatching condi-

tions. We measured slopes and intercepts by running a regression

weighed by the number of replications in each stimulus composi-

tion condition. This analysis assesses whether priming generates a

general decrease in processing time (an intercept effect) or if proc-

essing facilitation is affected by the stimulus’s length (a slope

effect, e.g., 4L stimuli beneﬁting from priming more than 1L stim-

uli, which should ﬂatten the slopes).

Slopes and intercepts were measured per participants and aver-

aged. The corresponding standard errors (standard error of the

intercept and standard error of the slope) were pooled (i.e., the

squared standard errors were averaged).

The average slope and intercept of the 0DlL and 1DlL condi-

tions for both priming conditions of Experiment 1 are presented in

the ﬁrst two rows of Table 2 as well as each statistic’s respective

pooled standard error (within parentheses). The Condition 1 col-

umn refers to the matching condition, and the Condition 2 column

refers to the mismatching condition.

Regarding intercepts, those for 1DlL trials are higher than those

for “same”trials in both priming conditions (a difference of

approximately 30 ms in both priming conditions; the two “same”

Table 2

Summary of Mean Slope (in ms per Letter, Noted ms/L) and Intercepts (in ms) for Both Priming Conditions of All Four Experiments for

All Participants

Experiment name Measure

Condition 1 Condition 2

Same 1D Same 1D

1. Case manipulation (match vs. mismatch) Slope 16.61 (1.31) 37.94 (2.75) 30.09 (1.22) 46.24 (3.92)

Slope ratio 2.28:1 (0.05:1) 1.54:1 (0.06:1)

Intercept 463.63 (3.57) 493.40 (5.80) 461.20 (3.31) 489.75 (8.28)

Intercept difference 29.77 (6.81) 28.55 (8.92)

2. Font manipulation (match vs. mismatch) Slope 25.11 (1.39) 50.46 (4.02) 36.39 (1.34) 42.55 (3.40)

Slope ratio 2.04:1 (0.04:1) 1.17:1 (0.08:1)

Intercept 481.66 (3.80) 527.82 (8.27) 473.75 (3.67) 532.32 (7.24)

Intercept difference 46.16 (9.10) 58.57 (8.12)

3. Stimuli presentation (visual vs. auditory) Slope 19.81 (1.53) 19.63 (3.18) 145.77 (2.96) 142.17 (5.80)

Slope ratio 0.99:1 (0.18:1) 0.98:1 (0.05:1)

Intercept 370.19 (8.06) 490.34 (12.29) 482. 96 (4.14) 594.40 (6.63)

Intercept difference 120.15 (14.70) 111.44 (7.82)

4. LTM association (present vs. absent) Slope 21.80 (3.44) 29.01 (1.87) 34.17 (3.53) 27.50 (2.92)

Slope ratio 1.33:1 (0.13:1) 0.80:1 (0.18:1)

Intercept 499.81 (7.45) 573.52 (5.10) 590.92 (7.58) 663.45 (7.93)

Intercept difference 73.71 (9.02) 72.54 (10.97)

Note. D = difference; LTM = long-term memory. Measures were only recorded for the “same”and 1DlL conditions. In parentheses are the standard error

of the estimates. In the leftmost column are the experiment names as well as the names of the priming conditions. Also presented within this table are the

slope ratios and intercept differences. The standard error of a ratio a/b is estimated from SE

a/b

(a/b)ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

SEa2=aþSEb2=b

p(Cousineau, 2020;Ku, 1966).

526 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

intercepts are almost identical, as are the two 1D intercepts).

Regarding slopes, “same”slopes are shallower than those for “dif-

ferent”in both priming conditions. However, there is a strong

slope decrease in the “same”matching condition in comparison to

its mismatching counterpart. The 1D-to-same slope ratio goes

down from 2.28:1 (38 ms/L vs. 17 ms/L) in the matching condition

to 1.54:1 (46 ms/L to 30 ms/L) in the mismatching condition.

In sum, for this ﬁrst experiment, the fast-same effect is linked

solely to a change in slope between priming conditions, and not a

change in intercept. Regarding “different”responses, slopes remained

relatively unchanged (as noted, there was no signiﬁcant interaction in

the 10 32 repeated-measures ANOVA for “different”RTs), and

intercepts were nearly identical across priming conditions.

Discussion

In this experiment, we observed the fan-out effect as well a fast-

same effect in the matching condition, replicating past results.

Furthermore, results were independent from the speciﬁc letter case

used. This conﬁrms once more that same-different task results

are not stimulus speciﬁc but are robust to changes in materials.

More importantly, we saw that a visual match is required to observe

the fast-same phenomenon, a ﬁnding akin to Bamber’s (1972) results

that this experiment aimed to conceptually replicate. The RTs for

“same”responses were severely attenuated in the mismatching

condition (a decrease of approximately .3 in terms of the Hedges’

gwhen comparing the overall means of “same”and “different”).

We have dubbed this relative slowdown the attenuated-same

effect. Meanwhile, “different”results were weakly affected by pri-

ming (the 10 32 ANOVA revealed a statistically signiﬁcant dif-

ference between the RTs of both priming conditions, albeit with a

small raw effect size). These results further highlight the necessity

of a visual match to observe fast-same responses and support the

priming hypothesis posited in the introduction. Furthermore, accu-

racy penalties are not observed when priming was altered. There

was a small accuracy beneﬁt in the matching 0D4L condition

compared to its mismatching counterpart, resulting in a signiﬁcant

interaction between the stimulus composition conditions and the

priming conditions; the accuracies for “different”showed no dif-

ference across priming conditions.

Slopes and intercepts offered additional information regarding

the possible mechanism at play. The slope is shallowest for

“same”responses in the matching condition, congruent with the

priming hypothesis that “same”conditions should be the only ones

beneﬁting from residual activation; the shallow slope also ﬁts the

expectations laid out by the fan-out effect noted by Sternberg

(1998). In the mismatching condition, the slope for “same”

responses is about twice as steep (by a factor of 1.82) and resem-

bles the slope of the 1D stimulus composition conditions. Finally,

the intercepts of “same”and “different”remained relatively

unchanged between priming conditions, being smaller for “same”

than for “different”in both cases.

As seen by these analyses, there are more similarities between

priming conditions than dissimilarities. The only major changes

between these conditions stem from the mean RT and the mean

RT slopes for “same”decisions. Appendix A further examines

standard deviation and skewness across priming conditions but

found no other notable differences. These results prompt us to

posit that there is an identical underlying decision-making process

for both visually matching and visually mismatching pairs of stim-

uli. The fast-same effect could therefore be due to a facilitating

factor in the process chain, not a different processing mechanism

altogether.

While this letter case manipulation experiment supports the pri-

ming hypothesis, it is unclear whether the observed results are

generalizable to other stimuli or to mismatches of a less extreme

magnitude. Uppercase and lowercase variants of a letter can be

visually different from one another and may even share no com-

mon features. This high level of discrepancy may be the cause of

the attenuated-same effect and also the cause of the small, but sig-

niﬁcant, difference between priming conditions for “different”

RTs. We therefore replicated this task in Experiment 2 with visual

mismatches of a smaller magnitude.

Experiment 2: Font Manipulation

To see if the fast-same attenuation found in Experiment 1 was

only caused by important visual mismatches, another same-differ-

ent experiment was carried out with subtler visual changes. In this

experiment, only the stimuli’s letters’font and typeface were var-

ied in the mismatching condition (both S

and S

were presented

as uppercase letters). As the visual attributes manipulated in the

present experiment are minor, results of this task will indicate

whether the fast-same effect can be attenuated in a gradual manner

or if it follows an all-or-naught rule.

Method

Participants were 20 new consenting adults aged 18 to 30 with

normal or corrected vision. Participants gave written and verbal

consent to participate in the experiment and were all informed of

the experiment’s procedure as well as the protocol and ethical

rules of the University of Ottawa (REB Certiﬁcate H10-16-14).

Finally, all participants were compensated $10 CAD for their

time.

The stimuli, procedure, and experimental design for this experi-

ment were identical to Experiment 1 except for what follows:

Rather than using matching/mismatching letter cases, we used

matching/mismatching font and typefaces on uppercase conso-

nants. For half of all trials, stimuli were shown in a Roman (i.e.,

nonitalic) Arial font (e.g., JCVD). In the second half, letters were

shown using an italic Bodoni MT font (e.g., JCVD). The Arial

font was selected as it does not include serifs (the small lines at

the end of a stroke), whereas the Bodoni MT font does include ser-

ifs, adding additional minute mismatches between stimuli. Once

more, the visually matching condition had an identical font and

typeface between S

and S

. The mismatch condition, exactly like

Experiment 1, presented visual mismatches between S

and S

Again, participants were instructed to pay no attention to the visual

nature of the stimuli and to make their “same”and “different”

decisions solely on the letters’identity.

Results

Screening of the Data

We gathered data from 15,360 total trials (768 trials per partici-

pant 320 participants), from which we removed 19 RTs below

200 ms, 59 RTs above 2,500 ms, and 37 nonanswers. Only correct

RESIDUAL ACTIVATION PRIMING 527

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

responses were retained for RT analyses, for a total of 14,521 tri-

als; there were 724 errors in total (5.0% error rate). Only a single

participant made ﬁve mistakes in a row once, which triggered the

warning message.

Effect of Font Differences

To ensure that both matching conditions and both mismatching

conditions were combinable, we performed two 14 32 repeated-

measures ANOVAs on the mean RTs using the 14 stimulus com-

position conditions as a factor and the two font/typeface variants

as the other factor (Arial/Roman or Bodoni/Italic for S

). In the

ﬁrst ANOVA comparing the font-matching conditions, the effect

of font was nonsigniﬁcant, F(1, 19) = .056, p= .815, h

= .003.

There was a main effect of stimulus composition, F(13, 247) =

17.917, p,.001, h

= .485, and no signiﬁcant interaction

between both factors, F(13, 247) = .544, p= .896, h

= .028. For

the ANOVA contrasting the mean RTs of the font-mismatching

conditions, the effect of font was nonsigniﬁcant, F(1, 19) = .513,

p= .483, h

= .026. There was a signiﬁcant effect of stimulus

composition, F(13, 221) = 14.953, p,.001, h

= .440, and a sig-

niﬁcant interaction between factors, F(13, 247) = 2.257, p= .008,

= .106. Although the interaction is signiﬁcant, it is mostly at-

tributable to a single discrepancy between the two 1D4L condi-

tions where the mean RT was 750.9 ms for the Arial/Roman font

and 647.6 ms for the Bodoni/Italic font. We decided to merge both

font-matching conditions and both font-mismatching conditions as

we had no theoretical interest in this interaction (the effect size is

also rather small).

Mean Response Times and Accuracy

Mean RT and accuracy rates for each priming condition (match-

ing and mismatching) are presented in Figure 3 in the same format

as Figure 2. As shown, the mean RTs for “different”are inﬂuenced

by stimulus composition in the usual manner (the fan-out effect is

observed here as well). The overall mean RT for “different”in the

matching condition is 602 ms, whereas it is 597 ms in the mis-

matching condition, a difference of 5 ms (with a 95% CI of 6

9.49 ms), this difference between overall mean “different”RTs is

nonsigniﬁcant, F(1, 19) = 1.431, p= .246, h

= .070. A 10 32

repeated-measures ANOVA on the mean RTs of the 10 “different”

stimulus composition conditions and the two priming conditions

showed that there is no effect of priming, F(1, 19) = 1.320, p=

.265, h

= .065, a main effect of stimulus composition, F(9, 171) =

23.283, p,.001, h

= .551, and a nonsigniﬁcant interaction

between both factors, F(9, 171) = .672, p= .734, h

= .034.

Regarding “same”trials, we see a fast-same effect in the match-

ing condition and an attenuated-same effect in the mismatching

condition. Overall, mean RTs for “same”responses in the match-

ing condition is 542 ms compared to 566 ms in the mismatching

condition, a difference of 24 ms (with a 95% CI of 69.27 ms),

this difference between overall mean “same”RTs is signiﬁcant, F

(1,19) = 24.902, p,.001, h

= .567.

Figure 3

Mean Response Time (RT; Top Panels) and Mean Accuracy (Bottom Panels) Results for

Experiment 2

Note. In the left column are the results for the matching condition, and in the right column are the results for

the mismatching condition. Error bars denote the correlation- and difference-adjusted 95% conﬁdence interval

of the mean (Baguley, 2012). D = differences. See the online article for the color version of this ﬁgure.

528 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

A432 repeated-measures ANOVA on the mean RTs of the four

“same”conditions and the two priming conditions showed an effect

of priming, F(1, 19) = 24.898, p,.001, h

= .567, an effect of stim-

ulus composition F(3, 57) = 48.175, p ,.001, h

= .717, and a sig-

niﬁcant interaction between factors, F(3, 57) = 3.027, p= .037, h

.137. Again, much like Experiment 1, the interaction points to a slope

change between both matching and mismatching “same”conditions.

There was a strong fast-same effect in the matching condition (a

Hedges’gof .57 between overall “same”and overall “different”

mean RTs with a 95% CI of [.41, .80]), this difference between over-

all mean RTs is signiﬁcant, F(1, 19) = 73.961, p,.001, h

= .796.

The effect is attenuated in the mismatching condition (a Hedges’gof

.30 between overall “same”and overall “different”mean RTs with a

95% CI of [.14, .50]), this difference between overall mean RTs is

signiﬁcant, F(1, 19) = 14.133, p= .001, h

= .427.

Overall accuracies for “different”decisions are 94.5% and

95.0% for both matching and mismatching conditions, respec-

tively, and are 96.0% and 95.5% for “same”decisions, in the same

order. For “different”trials, a 10 32 repeated-measures ANOVA

on arcsine-transformed mean accuracy rates showed no effect of

priming, F(1, 19) = .177, p= .679, h

= .009, a main effect of

stimulus composition, F(9, 171) = 26.623, p,.001, h

= .584,

and a nonsigniﬁcant interaction between factors, F(9, 171) = .291,

p= .977, h

= .015. For “same”trials, a 4 32 repeated-measures

ANOVA on arcsine-transformed accuracy rates similarly showed

no effect of priming, F(1, 19) = 1.696, p= .208, h

= .082, a main

effect of stimulus composition, F(3, 57) = 3.570, p= .019, h

.158, and a nonsigniﬁcant interaction between factors, F(3, 57) =

1.955, p= .131, h

= .093.

RT Slopes and Intercepts

To see whether the attenuated slope effect observed in Experiment

1 was also present in this experiment, the overall slopes and inter-

cepts of the 1DlLand“same”conditions were measured in both pri-

ming conditions. The results of this analysis are presented in the

second section of Table 2 where Condition 1 refers to the matching

condition, and Condition 2 refers to mismatching condition.

As seen in Table 2, the intercepts for 1D are higher than those

of “same”regardless of the priming condition, which is an identi-

cal result to that of Experiment 1. This result suggests that there

may be a distinct baseline latency for “same”decisions relative to

“different”decisions. The slopes are also inﬂuenced by the pri-

ming condition: The 1D-to-same slope ratio for matching trials is

2.04:1 (50 ms/L vs. 25 ms/L), whereas it is 1.17:1 (43 ms/L vs. 36

ms/L) for mismatching trials. The mismatching “same”trials have

a slope that is 1.44 times steeper than the matching “same”trials,

as anticipated from the signiﬁcant interaction seen in the 4 32

ANOVA on mean RTs.

Discussion

Results from this experiment show that minute changes in visual

identity still led to the trends found in Experiment 1. Moreover,

RTs for mismatching “different”decisions are not signiﬁcantly

different from their matching counterparts and ﬁt the expected

fan-out effect in both priming conditions; the accuracy trends for

both matching and mismatching priming conditions also ﬁt what is

expected from the task and are similar to the results from Experi-

ment 1. More importantly, even if the mismatch condition presents

subtle changes in visual identity, the fast-same effect is still

severely attenuated as was the case in Experiment 1. In fact, there

is a difference of .27 between the Hedges’gof both priming con-

ditions in both experiments; the mean Hedges’gfor the matching

conditions are also nearly identical at .55 and .57 for Experiments

1 and 2, respectively. Compared to Experiment 1, the intercept dif-

ference increased by a magnitude of approximately 24 ms, and the

slope ratios decreased by approximately .30 in both the matching

and mismatching conditions. Thus, the beneﬁts gained from the

fast-same effect may not be solely restricted to a slope effect. We

return to these results in the “General Discussion”section.

This experiment shows that a very strong visual match—and the

priming beneﬁts associated with that factor—is necessary to obtain

the fast-same effect. Without its presence, “same”responses are

much slower (yet are still faster than the majority of “different”

RTs and are never slower than the 1DlL condition). We can thus

conclude that a change in priming, as minor/major as it may be,

results in an attenuation of “same”mean RTs. No other result

seems to be as affected by the priming condition—not the “differ-

ent”mean RTs, the accuracies of both decisions, nor the higher-

moment statistics as shown in Appendix A.

Experiment 3: Visual Versus Auditory Stimuli

In this third experiment, we explored priming at the higher lev-

els of the mental representation hierarchy by constraining its oper-

ation to the phonological level and up. By presenting the stimuli

aurally to participants, it is consequently possible to bypass any

inﬂuence that visual priming may have. If priming is indeed an all-

or-none effect, as suggested by the results of Experiment 2, we

should observe similar results between the auditory priming condi-

tion presented here and the mismatching conditions of Experi-

ments 1 and 2. As in the previous experiments, there is also a

control condition where visually matching stimuli are presented.

Method

We recruited 20 additional adults aged 18 to 30 with normal or

corrected vision for this experiment. All participants were

informed of the experiment’s procedure, as well as the protocol

and ethical rules of the University of Ottawa (REB Certiﬁcate

H10-16-14). Participants gave written and verbal consent to partic-

ipate in this task and were compensated $10 CAD for their time.

The experiment’s control condition had an identical procedure,

experimental design, and stimuli to those of Experiments 1 and 2

except for the following: In both priming conditions, an additional

stimulus consisting of underscores was presented between the ﬁxa-

tion point and S

. This stimulus was presented for 505 ms, and the

number of underscores presented on-screen matched the length of

so that the criterion string did not end unexpectedly when it

was presented aurally. In the auditory condition, S

was broken

down into its individual letters and pronounced serially (1 s per

letter; e.g., 4L stimuli took 4 s to present) through desktop speak-

ers. The presentation language was assigned nonrandomly based

on the participant’s primary language, either French or English

since the University of Ottawa is a bilingual university. There was

an example auditory stimulus prior to the testing phase to ensure

that participants could hear and identify the presented stimuli

in the task. The auditory stimuli were generated using a voice

RESIDUAL ACTIVATION PRIMING 529

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

synthesizer (Version 1.56 of DSpeech, with the voices of Micro-

soft Anna used for English stimuli and ScanSoft Virginie used for

French stimuli). For visual S

, all letters composing the string

were presented at once for 400 ms. In all trials, S

was presented

visually. Figure 4 presents a schematic of the updated procedure

used for this experiment.

Results

Screening of the Data

Data were gathered from 14,592 total trials (768 trials 319 par-

ticipants). One participant was removed from the analysis for hav-

ing very slow RTs (a mean of 1,238 ms). Of the remaining data,

we removed three trials with RTs below 200 ms, 86 trials with

RTs above 2,500 ms, and 11 nonanswers. For RT analyses, we ﬁl-

tered out erroneous trials, resulting in a total of 13,690 trials. There

were 802 total errors (5.5% error rate). Of the 19 participants

retained for analysis, none made ﬁve mistakes in a row.

Mean Response Times and Accuracy

Mean RT and accuracy rates for each of the priming conditions

(auditory vs. visual) are presented in Figure 5. Note that for visual

stimuli, we plotted the results with two different scales, one that

matches the scale used in Experiments 1 and 2 and one enlarged to

ease comparisons with the auditory stimulus condition.

As seen in Figure 5, auditory trials lead to a considerable slow-

down in response times. This is possibly caused by the sequential

presentation of S

, which might have induced a more serial inspec-

tion of S

. Despite this major increase in RT, “different”conditions

follow the usual fan-out effect, and the fast-same phenomenon is

not observed. Again, “same”RTs are still below the 1D line, indi-

cating that all “same”conditions remain faster than the slowest

“different”condition even when a phonological mental representa-

tion is compared to a visually presented stimulus.

For “different”responses, overall mean RT is 610 ms for visual

trials and 748 ms for auditory trials, a difference of 138 ms (with a

95% CI of 641.38 ms); this difference between overall mean RTs

is signiﬁcant, F(1, 18) = 83.743, p,.001, h

= .823. Because of

this large slowdown seen in the auditory condition, all compari-

sons involving the priming condition will be signiﬁcant as well.

A1032 repeated-measures ANOVA on the mean RTs of “dif-

ferent”trials involving the 10 stimulus composition conditions and

the two priming conditions showed a signiﬁcant effect of priming,

F(1, 18) = 83.131, p,.001, h

= .822, a signiﬁcant effect of stim-

ulus composition, F(3, 54) = 74.724, p,.001, h

= .806, and a

signiﬁcant interaction between factors, F(3, 54) = 56.760, p,

.001, h

= .759.

Concerning “same”trials, overall mean RT is 532 ms for visual

trials and 731 ms for auditory trials, a difference of 198 ms (with a

95% CI of 645.54 ms); this difference between overall mean RTs

is signiﬁcant, F(1, 18) = 49.155, p,.001, h

= .732. A 4 32

repeated-measures ANOVA on the RTs of “same”trials involving

the four stimulus composition conditions and the two priming con-

ditions showed a signiﬁcant effect of priming, F(1, 18) = 74.484,

p,.001, h

= .805, a signiﬁcant effect of stimulus composition,

Figure 4

Timeline of an Auditory Trial for Experiment 3

Note. S

denotes the auditory stimulus presented to participants (letters are vocalized serially in 1-s incre-

ments), and S

denotes the second stimulus to be visually presented to the participant. The underscore(s) is/are

presented immediately after the ﬁxation point and indicates the length of S

. As previously stated, feedback

is only given on errors and nonresponses. In the visual condition, S

is presented visually for 400 ms. In the

actual experiment, ER was the word “Error”and NR was the words “No response detected,”both shown in the

lower part of the display below S

. See the online article for the color version of this ﬁgure.

530 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

F(9, 162) = 40.451, p,.001, h

= .692, and a signiﬁcant interac-

tion between factors, F(9, 162) = 22.010, p,.001, h

= .550.

Between overall “same”and overall “different”mean RTs for vis-

ual trials, there is a Hedges’g of .79 (with a 95% CI of [.56, 1.12]);

this difference between overall mean RTs is signiﬁcant, F(1, 18) =

63.315, p,.001, h

= .779. For auditory trials, there is a Hedges’g

of .10 between overall “same”and overall “different”mean RTs

(with a 95% CI of [.08, .29]); this difference between overall mean

RTs is nonsigniﬁcant, F(1, 18) = 1.278, p= .273, h

= .066.

Participants are 95.4% and 94.0% accurate for visual “different”

and “same”conditions, respectively, and 94.8% and 93.7% accu-

rate for auditory “different”and “same”conditions, respectively.

This is congruent with what was found in previous experiments. A

10 32 repeated-measures ANOVA on the arcsine-transformed

mean accuracy rates of “different”trials showed no signiﬁcant

effect of priming, F(1, 18) = 2.779, p= .113, h

= .134, a signiﬁ-

cant effect of stimulus composition, F(9, 62) = 23.382, p,.001,

= .565, and a signiﬁcant interaction between factors, F(9, 62) =

2.639, p= .007, h

= .128. The interaction comes from easier trials

(the four all-different conditions: 1D1L, 2D2L, 3D3L, and 4D4L)

that are more accurate in the auditory condition than in the visual

condition (97.5% vs. 96.2%, respectively). A 4 32 repeated-

measures ANOVA on the arcsine-transformed mean accuracy

rates of “same”trials showed a signiﬁcant effect of priming (albeit

with a small raw effect size of þ.3%), F(1, 18) = 6.855, p= .017,

= .276, a nonsigniﬁcant effect of stimulus composition, F(3,

54) = 1.849, p= .149, h

= .093, and a nonsigniﬁcant interaction

between factors, F(3, 54) = 1.726, p= .173, h

= .087.

RT Slopes and Intercepts

Mean slopes and intercepts are presented in the third section of

Table 2. Condition 1 refers to visual trials, and Condition 2 refers to

auditory trials. With regard to the visual condition, the intercept for

“same”is lower than the intercept for 1D by a magnitude of approxi-

mately 120 ms. This is roughly 90 ms more than what was observed

in Experiment 1 and approximately 70 ms more than what was

observed in Experiment 2. However, the slopes are almost at a 1:1

ratio (20 ms/L:20 ms/L for “same”onto 1D). This lack of difference

between slopes contradicts what we observed for fast-same

responses so far in Experiments 1 and 2 where the ratio was near

2:1. Nevertheless, because there is an encoding modality switch in

this paradigm, it is unsurprising that not all results have carried over.

As for auditory stimuli, the intercept for “same”responses is

still quicker than that for 1DlL responses. Furthermore, slopes fol-

low a 1:1 ratio, which is more in line with the attenuated-same

effect we have seen so far. However, we observe very steep slopes

for both conditions compared to the other experiments (about 140

ms/L here vs. about 40 ms/L elsewhere), which indicates that stim-

uli are processed at a much slower rate. This could be due to the

Figure 5

Mean Response Time (RT; Top Panels; Note Different Scaling Between Left and Middle Panels) and Mean Accuracy (Bottom Panels)

Results for Experiment 3

Note. In the top-left and top-middle panels are the results for visually presented trials (one has the same vertical axis scaling as Experiments 1 and 2;

the other has the same scaling as auditory trials), and in the top-right panel are the results for the auditory trials. Error bars denote the correlation- and

difference-adjusted 95% conﬁdence interval of the mean (Baguley, 2012). D = differences. See the online article for the color version of this ﬁgure.

RESIDUAL ACTIVATION PRIMING 531

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

serial nature of S

that might also induce a more serial mental rep-

resentation/comparison process. The latter would also explain why

accuracy is improved in the all-different conditions.

Discussion

In this task, we presented S

to participants through two different

sensory modalities. For the visual condition, results concur with the

visually matching conditions of Experiments 1 and 2. The visual con-

dition’s mean RT and mean accuracy, as well as intercept measures,

follow the same trend and lead to the same conclusion about the fast-

same phenomenon as previous experiments. The only two differen-

ces between this experiment and previous ones are that (a) the fast-

same effect here is expressed by a pronounced intercept reduction

rather than by a change in slope and (b) the RT slope ratio is nearly

1:1 for the control condition as opposed to the 2:1 ratio previously

seen. As mentioned, these differences could be due to a switch in

encoding modality. Still, mean “same”RTs remain faster than those

of the slowest “different”condition, and the Hedges’gbetween over-

all “same”and overall “different”mean RTs is larger here than it is

for the control conditions of Experiments 1 and 2. Regarding audi-

tory trials, they further support the hypothesis that visually matching

information is necessary for a fast-same effect to occur.

Experiment 4: Long-Term Memory Associations

In this last experiment, the role of priming within the same-dif-

ferent task is further explored by comparing results from bottom-

up activation (as is the case with the three experiments presented

above) with top-down stimulation (stimuli held in LTM). In the

present experiment, and for both priming conditions, participants

were required to memorize four stimuli of speciﬁc lengths (a sin-

gle stimulus was memorized for each of the task’s four possible

string lengths). In the “letters shown”condition (S

-shown hence-

forth), letters were presented directly to the participants as usual.

In the LTM retrieval condition (S

-absent henceforth), the only

visual stimulus presented on-screen was a visual cue, composed of

underscores, indicating the stimulus’s overall length. From this

cue, participants had to access their LTM to retrieve the memo-

rized string and compare it to S

. In this priming condition, neither

visual representations nor phonological representations are stimu-

lated; the only possible mental representation comes from the

association existing between the number of underscores presented

on-screen and the memorized strings associated to each stimulus

length. Considering the results found thus far, we believed that a

fast-same effect would not be observed in the S

-absent condition

as priming cannot operate at the visual level, while the effect will

be seen in the S

-shown manipulation (which is much closer to a

standard same-different task). We also predicted that overall deci-

sions will be more accurate in the S

-shown condition due to the

very small number of stimuli presented repeatedly to each partici-

pant, possibly yielding training effects (see Walker & Cousineau,

2019, for a similar manipulation with extended training). Addi-

tionally, due to these training effects and the atypical stimuli pre-

sentation method, it is uncertain whether the typically observed

mean RT and accuracy patterns will be elicited from this task as it

has now shifted from a comparison task to an LTM retrieval þ

comparison task.

Method

All stimuli and procedures and the experimental design for

the control condition were identical to Experiment 1 with the

exception that letters composing S

were not chosen randomly on

every trial (see below). Participants were 21 adults aged 18 to 30

with normal or corrected vision. All participants were informed of

the experiment’s procedure, as well as the protocol and ethical

rules of the University of Ottawa (REB Certiﬁcate H10-16-14).

All participants gave written and verbal consent to participate in

this task and were compensated $10 CAD for their time.

In a preliminary phase, participants were asked to memorize

four separate strings of letters (randomly generated for each partic-

ipant), each corresponding to a speciﬁc string length. For example,

if the prospective four-letter string was generated as “JCVD,”all

4L S

in the experiment would be presented as “____”or “JCV

D”; if the prospective two-letter string was “CN,”all 2L S

within

the experiment would be presented as “__”or “CN.”To ensure

that the four randomly generated strings were learned adequately,

participants were subjected to a memory test prior to testing: One

to four underscore symbols were randomly shown on-screen, and

the participant had to manually type the appropriate memorized

string. This preliminary memory test was repeated until partici-

pants could recall all four strings twice in a row without error.

The present experimental design was identical to Experiments 1

and 2. However, rather than randomly generating a different S

every trial, only the four memorized strings were used. On half of the

trials, S

was presented as underscores corresponding to the criterion

stimulus’s length, and on the other half, the actual memorized stimu-

lus was visually presented. Therefore, there were only four criterion

stimuli in the entire task (generated randomly for each participant).

Results

Screening of the Data

We gathered data from 14,592 total trials (768 trials 319 par-

ticipants). Twenty-one participants ran the experiment, but only 19

were retained for analysis; one participant was replaced after post-

testing accuracy analyses showed near chance-level accuracies in

the task (52% in a task that typically yields 95% and over).

Another participant was removed during the analysis period for

having unusually slow RTs (a mean of 1,128 ms). Of the remain-

ing trials, we removed 17 RTs below 200 ms, 98 RTs above 2,500

ms, and 22 nonanswers. For RT analyses, we ﬁltered out erroneous

trials for a total of 13,463 trials. There were 1,000 total errors

(6.9% error rate), slightly more than in Experiments 1 to 3. No

participant among those retained made ﬁve mistakes in a row.

Mean Response Times and Accuracy

Mean RT and accuracy rates for each of the priming conditions

-shown and S

-absent) are presented in Figure 6. Note the slightly

extended vertical axes for RT panels in comparison to those of previ-

ous experiments to encapsulate the full conﬁdence intervals.

For “different”responses, overall mean RT is 609 ms and 686

ms for S1-shown and S1-absent trials, respectively (a slowdown of

78 ms with a 95% CI of 625.41 ms); this difference between

overall mean RTs is signiﬁcant, F(1, 18) = 41.529, p,.001,

= .698. A 10 32 repeated-measures ANOVA on the mean

532 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

RTs of “different”decisions showed that there is a signiﬁcant

effect

of priming, F(1, 18) = 42.057, p,.001, h

= .700, a signiﬁcant

effect of stimulus composition, F(9, 162) = 10.913, p,.001, h

.377, and no signiﬁcant interaction between factors, F(9, 162) =

1.319, p= .231, h

= .068.

For “same”decisions, overall mean RT is 543 ms and 661 ms for

S1-shown and S1-absent trials, respectively (a slowdown of 119 ms

with a 95% CI of 642.26 ms); this difference between overall

mean RTs is signiﬁcant, F(1, 18) = 34.976, p,.001, h

=.660.A

432 repeated-measures ANOVA on the mean RTs of “same”

decisions showed that there is a signiﬁcant effect of priming, F(1,

18) = 34.798, p,.001, h

= .659, a signiﬁcant effect of stimulus

composition, F(3, 54) = 4.667, p=.006,h

= .206, and a signiﬁcant

interaction between factors, F(3, 54) = 6.261, p=.001,h

=.258.

This interaction is due to the recurved slope found between the 0D3L

and 0D4L stimulus composition conditions of S

-absent stimuli.

For S1-shown trials, there is a fast-same effect when comparing

overall mean “same”RTs to overall mean “different”RTs as indi-

cated by a Hedges’g of .39 (with a 95% CI of [.23, .61]); this dif-

ference between overall mean RTs is signiﬁcant, F(1, 18) =

16.814, p = .001, h

= .483. For the S1-absent condition, there is

an attenuated-same effect as indicated by a Hedges’g of .11 (with

a 95% CI of [.01, .24]); this difference between overall mean

RTs is nonsigniﬁcant, F(1, 18) = 3.446, p= .080, h

= .161. In

fact, in this priming condition, “same”RTs are slower than all but

the 1DlL and 3D4L stimulus composition conditions, allowing us

to conclude that the fast-same effect is considerably attenuated;

-absent trials are slower than the visually mismatching trials of

Experiments 1 and 2, suggesting that memory retrieval added

some additional latency not seen in the S

-shown condition.

In S

-shown trials, participants are 96.1% and 96.4% accurate

for “different”and “same”decisions, respectively. In S

-absent tri-

als, participants are 94.9% and 93.5% accurate for “different”and

“same”decisions, respectively. The S

-absent “same”conditions

are less accurate than the S

-shown “same”conditions by 2.9%,

the most important difference seen in all experiments so far. This

result is congruent with RT results: S

retrieval might not be fully

performed so that latency is increased, and errors are a bit more

frequent. A 10 32 repeated-measures ANOVA on arcsine-trans-

formed mean accuracy rates for “different”trials showed no effect

of priming, F(1, 18) = 1.324, p= .265, h

= .069, a signiﬁcant

effect of stimulus composition, F(9, 162) = 11.058, p,.001, h

.381, and a nonsigniﬁcant interaction between factors, F(9, 162) =

1.011, p= .434, h

= .053. A 4 32 repeated-measures ANOVA

on arcsine-transformed mean accuracy rates for “same”trials

showed that there is a signiﬁcant effect of priming, F(1, 18) =

12.136, p= .003, h

= .403, no effect of stimulus composition, F

(3, 54) = .050, p= .985, h

= .003, and a nonsigniﬁcant interaction

between factors, F(3, 54) = .104, p= .957, h

= .006. The lack

Figure 6

Mean Response Time (RT; Top Panels; Note the Slightly Extended Vertical Axis in Comparison to

Figures 2,3, and 5) and Mean Accuracy (Bottom Panels) Results for Experiment 4

Note. In the left panels are the results for the S

-shown condition, and in the right panels are the results for

the S

-absent condition. Error bars denote the correlation- and difference-adjusted 95% conﬁdence interval of

the mean (Baguley, 2012). D = differences. See the online article for the color version of this ﬁgure.

RESIDUAL ACTIVATION PRIMING 533

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

of a stimulus composition effect indicates that the accuracy of

“same”conditions is ﬂat with respect to string length. Previous

experiments had nearly ﬂat results, so this nonsigniﬁcant effect is

not unexpected.

RT Slopes and Intercepts

As for slope analyses, results are presented in the fourth and last

section of Table 2. Condition 1 refers to S

-shown stimuli, while

Condition 2 refers to S

-absent stimuli. For S

-shown stimuli, the

intercept for 1D is higher than the intercept for “same,”typical of

the results shown thus far. There is a moderate decrease in slope

(1.3:1) and a fairly important increase in intercept (approximately

70 ms). As for the S

-absent condition, the slope ratio is below 1:1

at .80:1, and the difference between intercepts is again approxi-

mately 70 ms. However, note in Figure 6 that the mean RT of

1D3L and 1D4L trials and the mean RT of 0D3L and 0D4L trials

in the S

-absent condition ﬂatten out, and the slope between those

conditions becomes negative—this artifact is likely the factor

affecting the overall measures of slope. Therefore, the sub-1:1 ra-

tio for slopes may be biased downward, even if the regression is

weighed by the number of replications per condition. Here, it

seems that decision latency for “same”RTs in the S

-absent condi-

tion is tied to a change in slope rather than a change in intercept as

was observed in Experiments 1 and 2.

Discussion

RTs in the S

-shown condition are similar to the visually match-

ing conditions of Experiments 1 and 2 and the visually presented

stimuli of Experiment 3. In fact, results of this experiment are

nearly identical to the ﬁrst two experiments, other than the ﬂatter

slope for 1D that induced a reduced slope ratio of approximately

1.3:1 rather than the approximate 2:1 ratio observed in those

experiments. This might be attributed to the possible training

effects of S

: As participants are repeatedly shown the same four

criterion stimuli, the mismatching letters in S

might be more sa-

lient relative to the primed letters due to an improved signal-to-

noise ratio. Regarding S

-absent trials, the overall pattern for “dif-

ferent”results matches the fan-out effect well, with the exception

of the slope ﬂattening out between the 0D3L-0D4L and 1D3L-

1D4L stimulus composition conditions. Yet mean “different”RTs

for S

-absent trials are all considerably slower than their S

-shown

counterparts. As the training effects are identical across priming

conditions, the differences between these experimental manipula-

tions are either due to (a) the presence of the target stimulus within

the S

-shown trials—training might magnify the priming beneﬁt,

which further highlights the fact that visually identical stimuli are

required to observe the fast-same effect—or (b) an imperfect

memory retrieval—the 400-ms interstimulus interval might be too

short to fully retrieve the criterion stimulus from LTM.

In this variant of the same-different task, no priming beneﬁts

from sensory modalities were afforded in half of the trials because

no visual or phonological information was presented to partici-

pants. In these trials, the fast-same effect was attenuated, and the

slope for “same”RTs was much closer to the slope of the 1DlL

conditions. Consequently, results show that visual information is

not necessary to complete the task but is necessary for the pres-

ence of a fast-same effect.

General Discussion

In all experiments, we altered priming between pairs of stimuli

for four same-different tasks that included one variant where S

and S

could be visually matching or mismatching using different

letter cases, a second variant where S

and S

could be visually

matching or mismatching using different fonts and typefaces, a

third variant where S

presentation was limited to phonology on

half of the trials, and a fourth variant where S

was not presented

at all on half of the trials. Throughout these experiments, when pri-

ming was not altered (the “control”conditions: matching trials,

visually presented trials, and S

-shown trials), RT and accuracy

for all “same”and “different”conditions follow the same trends

established by the literature (see Sternberg, 1998, for a thorough

overview of the expected results), and we observe the presence of

a fast-same effect (the four experiments have a mean Hedges’gof

.55, .57, .79, and .39, respectively, all favoring overall mean

“same”RTs relative to overall mean “different”RTs). Trials

intended to cancel priming (the “experimental”conditions: mis-

matching trials, auditory trials, and S

-absent trials) tell the same

story: Accuracy and RT for “different”conditions are similar to

their “control”counterparts, whereas the rapidity of “same”deci-

sions is severely attenuated (the four experiments respectively

have a mean gof .28, .30, .10, and .11 between the overall mean

RTs of “same”and “different”). Thus, the fast-same effect is not

unequivocally characterized by a lower slope (contrary to the

results of Bamber, 1969;Sternberg, 1998;Taylor, 1976a). While

this was the case in Experiments 1 and 2 (1D-to-same slope ratios

of 2.28:1 and 2.04:1 for Experiments 1 and 2, respectively), in

Experiment 3, the slopes were near equal with a slope ratio of

.99:1. In this experiment, it was the intercepts of 1D and “same”

that varied between priming conditions (the intercept of 1D was

111.44 ms higher than the intercept for “same”in this experiment).

For Experiment 4, the slope ratio was in between those found in

previous experiments at 1.33:1, as was the intercept difference

between the 1D condition relative to the “same”condition at 73.71

ms. Nevertheless, intercept measures for “same”are always faster

than those for “different.”This suggests that there may be an in-

herent bias for “same”decisions that is unaffected by priming. As

Huber (2008),Eviatar et al. (1994), and Lupker et al. (2015) have

noted, response biases are not limited to the perceptual stage and

could also operate at the response selection stage, leaving the dis-

crepancies between intercepts of “same”and 1D unaffected by the

priming manipulation.

In all four experiments, attenuated-same responses do not come

with a severe accuracy penalty as they are just a little lower than

their fast-same counterparts. Statistical tests found signiﬁcant dif-

ferences between arcsine-transformed accuracy rates in Experi-

ments 1, 3, and 4 but found no such signiﬁcance in Experiment 2;

the actual differences between priming conditions are rather small

(a raw effect size of 1.4%, .5%, .3%, and .9% between priming

conditions of “same”conditions for Experiments 1, 2, 3, and 4,

respectively).

It is our belief that the attenuated-same effect is equivalent to a

“same”decision with no added visual priming beneﬁts and that a vis-

ual match between S

and S

is necessary to obtain fast-same

responses. The equivalence here is meant in a strong sense: These two

effects reﬂect—process-wise—thesamebehavior.InAppendix A,we

examined higher statistical moments and found no differences between

534 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the fast-same and attenuated-same conditions (for Experiments 1 and

2). A change in processing mode when priming is removed would cer-

tainly entail an altered standard deviation or skew.

The change in processing rates between fast-same and attenuated-

same RTs (seen by changes in slopes) offers insight into the nature

of the fast-same phenomenon. It seems that “same”responses elicit

some sort of “pop-out”effect (Treisman, 1985;Wolfe, 1994)when

is primed. The pop-out effect is often deﬁned as having partici-

pants quickly identify a target in a set of foils. This idea might apply

here as “same”responses seem to be quite evident to the participant,

whereas “different”responses are more thoroughly processed. Thus,

one could maintain an identical analytical model for primed and non-

primed stimuli when the process is supplemented with a front-loaded

matching detector. In this view, matches become the relevant attrib-

utes attached to the perceived stimuli. By contrast, in the processing

facilitation view, when the stimulus (S

) matches the template (S

the decision is processed analytically with a higher (preactivated)

starting point. If the template does not match, no beneﬁt is given to

the accrual of “same.”In either proposal, the underlying decision

mechanism behind both fast- and attenuated-same responses are

identical, a proposal in same-different task research that contrasts

with, for example, Bamber's (1969) dual-process mechanism.

Potter et al. (2018) considered this possibility through their use

of the diffusion race model (Logan et al., 2014). The fact that

model ﬁtting suggested no role for thresholds in explaining the

fast-same effect is congruent with our own analyses (Appendix B,

using EZ) and the results of others using a variety of models (e.g.,

Gomez et al., 2013;Goulet & Cousineau, 2020a;Groulx et al.,

2020). Thus, even if the intercepts for “same”are lower than those

for 1D, a strong consensus emerges that biases and strategic

adjustments are not the cause of the fast-same effect, a position

heralded by the works of Proctor et al. (1984). More work is

required to investigate the discrepancy between intercept measures

to decode why “same”intercepts are always lower than their “dif-

ferent”counterparts even when there is no priming between S

and S

Farell (1977), using a much longer criterion stimulus presenta-

tion duration (1.5 s), also reported an attenuated fast-same effect.

This result is also in line with this research’s priming hypothesis,

as well as Huber’s (2008) work where a reduction or even reversal

of the priming beneﬁts is found following long stimulus

presentations.

The front-loaded matching detector hypothesis is congruent

with the nROUSE model of residual activation (Huber, 2008;

Huber & O’Reilly, 2003) in which stimuli are compared and proc-

essed in a series of hierarchical levels. The lower, visual matching

level acts as a gatekeeper as to whether the stimuli can be fast-

tracked or not. This fast-tracked beneﬁt results in faster responses

that make “same”decisions “feel obvious”compared to “differ-

ent”decisions (an anecdotal comment made by many of the 81

total participants). In Potter et al. (2018), this fast-track idea is

called ﬂuency of processing. If the stimuli are visually mismatch-

ing, then they must go through all levels analytically as is the case

for all “different”decisions; residual activation can therefore not

operate nor beneﬁt the response. For “same”trials that are not

fast-tracked, string length is directly proportional to RT, as is the

case with many cognitive architectures (Townsend & Ashby,

1983), resulting in steeper slopes. An unanswered question in

these experiments concerns what would happen when S

and S

are only partially visually matching. Using Experiment 2 as an

example, we could have some letters unchanged between S

and

while others have an altered font, or with regard to Experiment

3, some letters could be presented aurally, and others presented

visually. Regarding Experiment 1, where some of the letters could

be of a different letter case, pilot testing showed that the task was

too difﬁcult. If the same occurs in all variants of the task where

visual matches can be partial, then it would argue against the

front-loaded match detector and against the view that matches are

additional attributes attached to the perceived stimuli. More inves-

tigation is required here.

Conclusion

Hampering priming abolishes the fast-same phenomenon with-

out substantial impact on the processing of “different”decisions;

the slopes for “same”responses were also attenuated, whereas no

major change was observed between “different”conditions across

priming manipulations. This shows not only that a change in

“same”latency has little impact on the accuracy patterns of either

decision but also that the fast-same phenomenon itself is either

fully operational or fully attenuated. Large changes (such as letter

case) and minor changes (such as font) were equally effective in

disrupting the fast-same effect. Thus, the fast-same effect might

indeed be all-or-nothing and tied to identity priming. This system-

atic approach to answering the research question offers an elegant

and parsimonious explanation of the fast-same effect, concurring

with the research of Nickerson (1978) and Proctor (1981).

While many other models could explain the fast-same phenom-

enon (of which a list was presented in the introduction), none has

been as diligently and successfully tested as the priming hypothe-

sis presented in this article. To create a viable model, future

research should expand the priming framework and, ideally,

instantiate its assumptions within a processing architecture to see

if it can reproduce the strong effects observed in these experi-

ments. Priming should be integrated into whatever models are

developed henceforth to account for the RT difference between

both fast- and attenuated-same priming conditions or be chal-

lenged by future research. Finally, the fact that “same”RTs can be

attenuated is sensible in the present model, but there is no explana-

tion on why they remain faster than the RTs of the slowest “differ-

ent”conditions. We believe this to be the next step in our research

—assessing stopping rule, measuring capacity as a function of

string length, and focusing on the response-selection stage may be

fruitful lines of investigation to make new progress.

Clearly, this study and decades of other research on the same-

different task show us that humans have a strong facility to detect

“sameness.”William James (1890) even emphatically wrote that

“The sense of sameness is the very keel and backbone of our

thinking [... and may be] the most important of all the features of

our mental structure”(James, 1890, pp. 459–460). Yet we do not

know why an animal would need such optimized sameness proc-

essing. What role does it play in our cognitive processes? Why do

we fast-track identical stimuli? An alternative view would be that

“same”decisions are not that fast—it is instead that “different”

decisions are slow, revealing the advanced analytical capacities of

our brain. In this view, lower animals would only be capable of

fast-same responses. Likewise, it seems that “matching”means

“absolutely identical”(consider the font changes in Experiment 2),

RESIDUAL ACTIVATION PRIMING 535

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

in which case the sameness detection would truly be a “same to-

ken”detection as opposed to the detection of objects of the same

type (Jackson & Buchanan, 2016;Swan & Wyble, 2014). The fact

that even these simple questions are still unanswered to this day

shows that the study of same-different abilities is still in its

infancy.

References

Baguley, T. (2012). Calculating and graphing within-subject confidence

intervals for ANOVA. Behavior Research Methods,44(1), 158–175.

https://doi.org/10.3758/s13428-011-0123-7

Bamber, D. (1969). Reaction times and error rates for “Same”and “Differ-

ent”judgments of multidimensional stimuli. Perception & Psychophy-

sics,6(3), 169–174. https://doi.org/10.3758/BF03210087

Bamber, D. (1972). Reaction times and error rates for judging nominal

identity of letter strings. Perception & Psychophysics,12(4), 321–326.

https://doi.org/10.3758/BF03207214

Bamber, D., & Paine, S. (1973). Information retrieval processes in

“Same”-“Different”judgments of letter strings. In S. Kornblum (Ed.),

Attention and performance IV. Academic Press.

Barth, M. (2019). ezDiffusion: The EZ-diffusion model as described by

Wagenmakers (R Package Version 0.1.0) [Computer software]. https://

rdrr.io/github/mariusbarth/ezDiffusion/

Beller, H. K. (1970). Parallel and serial stages in matching. Journal of Experi-

mental Psychology,84(2), 213–219. https://doi.org/10.1037/h0029065

Bertelson, P. (1961). Sequential redundancy and speed in a serial two-

choice responding task. The Quarterly Journal of Experimental Psychol-

ogy,13(2), 90–102. https://doi.org/10.1080/17470216108416478

Bindra, D., Donderi, D. C., & Nishisato, S. (1968). Decision latencies of

“Same”and “Different”judgments. Perception & Psychophysics,3(2),

121–136. https://doi.org/10.3758/BF03212780

Bindra, D., Williams, J. A., & Wise, J. S. (1965). Judgments of sameness

and difference: Experiments on decision time. Science,150(3703),

1625–1627. https://doi.org/10.1126/science.150.3703.1625

Cousineau, D. (2005). Confidence intervals in within-subjects designs: A

simpler solution to Loftus and Masson’s method. Tutorials in Quantita-

tive Methods for Psychology,1(1), 42–45. https://doi.org/10.20982/tqmp

.01.1.p042

Cousineau, D. (2017). Varieties of confidence intervals. Advances in Cog-

nitive Psychology,13(2), 140–155. https://doi.org/10.5709/acp-0214-z

Cousineau, D. (2020). How many decimals? Rounding descriptive and

inferential statistics based on measurement precision. Journal of Mathe-

matical Psychology,97, Article 102362. https://doi.org/10.1016/j.jmp

.2020.102362

Cousineau, D., & Larochelle, S. (2004). Visual-memory search: An inte-

grative perspective. Psychological Research,69(1-2), 77–105. https://

doi.org/10.1007/s00426-003-0170-5

Derks, P. L. (1972). Visual recognition of similarity and identity. Journal

of Experimental Psychology,95(1), 237–239. https://doi.org/10.1037/

h0033282

Donderi, D. C., & Zelnicker, D. (1969). Parallel processing in visual same-

different decisions. Perception & Psychophysics,5(4), 197–200. https://

doi.org/10.3758/BF03210537

Dyer, F. N. (1973). Same and different judgments for word-color pairs

with “irrelevant”words or colors: Evidence for word-code comparisons.

Journal of Experimental Psychology,98(1), 102–108. https://doi.org/10

.1037/h0034278

Egeth, H. E. (1966). Parallel versus serial processes in multidimensional

stimulus discrimination. Perception & Psychophysics,1(4), 245–252.

https://doi.org/10.3758/BF03207389

Eviatar, Z., Zaidel, E., & Wickens, T. (1994). Nominal and physical deci-

sion criteria in same-different judgments. Perception & Psychophysics,

56(1), 62–72. https://doi.org/10.3758/bf03211691

Farell, B. (1977). Encoding and comparisons in “Same”-“Different”judg-

ments [Unpublished doctoral dissertation]. McGill University.

Farell, B. (1985). Same-different judgments: A review of current contro-

versies in perceptual comparisons. Psychological Bulletin,98(3),

419–456. https://doi.org/10.1037/0033-2909.98.3.419

Gomez, P., Perea, M., & Ratcliff, R. (2013). A diffusion model account of

masked versus unmasked priming: Are they qualitatively different?

Journal of experimental psychology: Human perception and Perform-

ance,39(6), 1731–1740. https://doi.org/10.1037/a0032333

Goulet, M. A., & Cousineau, D. (2019). The power of replicated measures

to increase statistical power. Advances in Methods and Practices in

Psychological Science,2(3), 199–213. https://doi.org/10.1177/2515245

919849434

Goulet, M. A., & Cousineau, D. (2020a). The fast-same effect of an exclu-

sive-OR task. Journal of Experimental Psychology: Human Perception

and Performance,46(9), 991–1000. https://doi.org/10.1037/xhp0000783

Goulet, M. A., & Cousineau, D. (2020b). Sequential sampling models of

same-different data and how they explain the fast-same effect. Canadian

Journal of Experimental Psychology/Revue canadienne de psychologie

expérimentale,74(4), 284–301. https://doi.org/10.1037/cep0000197

Goulet-Pelletier, J.-C., & Cousineau, D. (2018). A review of effect sizes

and their confidence intervals, Part I: The Cohen’s d family. The Quanti-

tative Methods for Psychology,14(4), 242–265. https://doi.org/10

.20982/tqmp.14.4.p242

Groulx, J. T., Harding, B., & Cousineau, D., (2020). The EZ diffusion

model: An overview with derivation, software, and an application to the

Same-Different task. The Quantitative Methods for Psychology,16(2),

154–174. https://doi.org/10.20982/tqmp.16.2.p154

Harding, B., Goulet, M. A., Jolin, S., Tremblay, C., Villeneuve, S.-P., &

Durand, G. (2016). Systems factorial technology explained to humans.

The Quantitative Methods for Psychology,12(1), 39–56. https://doi.org/

10.20982/tqmp.12.1.p039

Harding, B., Goulet, M. A., Lauzon, S., & Cousineau, D. (2021). Redun-

dancy and the Same-Different task: An argument for coactivity in the

comparison paradigm [Manuscript submitted for publication]. École de

Psychologie, Université de Moncton

Harding, B., Tremblay, C., & Cousineau, D. (2014). Standard errors: A

review and evaluation of standard error estimators using Monte Carlo

simulations. The Quantitative Methods for Psychology,10(2), 107–123.

https://doi.org/10.20982/tqmp.10.2.p107

Heitz, R. P. (2014). The speed-accuracy tradeoff: History, physiology,

methodology, and behavior. Frontiers in Neuroscience,8, Article 150.

https://doi.org/10.3389/fnins.2014.00150

Henmon, V. A. C. (1911). The relation of the time of a judgment to its ac-

curacy. Psychological Review,18(3), 186–201. https://doi.org/10.1037/

h0074579

Hock, H. S. (1973). The effects of stimulus structure and familiarity on same-

different comparison. Perception & Psychophysics,14(3), 413–420. https://

doi.org/10.3758/BF03211176

Howell, P., & Stockdale, J. (1975). Memory and display search in binary classi-

fication reaction time. Perception & Psychophysics,18(6), 379–388. https://

doi.org/10.3758/BF03204109

Huber, D. E. (2008). Immediate priming and cognitive aftereffects. Journal

of Experimental Psychology: General,137(2), 324–347. https://doi.org/

10.1037/0096-3445.137.2.324

Huber, D. E., & O’Reilly, R. C. (2003). Persistence and accommodation in

short-term priming and other perceptual paradigms: Temporal segrega-

tion through synaptic depression. Cognitive Science,27, 403–430.

https://doi.org/10.1207/s15516709cog2703_4

536 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Jackson, A., & Buchanan, L. (2016). Survival of the grouped, or three’sa

crowd? Repetition blindness in groups of letters and words. Memory &

Cognition,44(2), 278–291. https://doi.org/10.3758/s13421-015-0556-9

Jacob, J., Breitmeyer, B. G., & Treviño, M. (2013). Tracking the first two

seconds: Three stages of visual information processing? Psychonomic

Bulletin & Review,20(6), 1114–1119. https://doi.org/10.3758/s13423

-013-0482-4

James, W. (1890). Principles of psychology (Vol. 1). Holt.

Kornblum, S. (1969). Sequential determinants of information processing in

serial and discrete choice reaction time. Psychological Review,76(2),

113–131. https://doi.org/10.1037/h0027245

Krueger, L. E. (1973). Effect of stimulus frequency on speed of “Same”-

”Different”judgments. In S. Kornblum (Ed.), Attention and perform-

ance V (pp. 497–506). Academic Press.

Krueger, L. E. (1978). A theory of perceptual matching. Psychological

Review,85(4), 278–304. https://doi.org/10.1037/0033-295X.85.4.278

Krueger, L. E. (1979). A model of unidimensional perceptual matching.

Journal of Experimental Psychology: Human Perception and Perform-

ance,5(2), 277–288. https://doi.org/10.1037/0096-1523.5.2.277

Krueger, L. E., & Shapiro, R. G. (1981). A reformulation of Proctor’s uni-

fied theory for matching-task phenomena. Psychological Review,88(6),

573–581. https://doi.org/10.1037/0033-295X.88.6.573

Ku, H. H. (1966). Notes on the use of propagation of error formulas. Jour-

nal of Research of the National Bureau of Standards C: Engineering

and Instrumentation,70C(4), 263–273. https://doi.org/10.6028/jres.070C

.025

Link, S. W., & Tindall, A. D. (1971). Speed and accuracy in comparative

judgments of line length. Perception & Psychophysics,9(3), 284–288.

https://doi.org/10.3758/BF03212649

Logan, G. D., Van Zandt, T., Verbruggen, F., & Wagenmakers, E.-J.

(2014). On the ability to inhibit thought and action: General and special

theories of an act of control. Psychological Review,121(1), 66–95.

https://doi.org/10.1037/a0035230

Lupker, S. J., Nakayama, M., & Perea, M. (2015). Is there phonologically

based priming in the same-different task? Evidence from Japanese-Eng-

lish bilinguals. Journal of Experimental Psychology: Human Perception

and Performance,41(5), 1281–1299. https://doi.org/10.1037/xhp0000087

Megreya, A. M., & Burton, A. M. (2006). Unfamiliar faces are not faces:

Evidence from a matching task. Memory & Cognition,34(4), 865–876.

https://doi.org/10.3758/bf03193433

Miller, J. (1982). Divided attention: Evidence for coactivation with redun-

dant signals. Cognitive Psychology,14(2), 247–279. https://doi.org/10

.1016/0010-0285(82)90010-X

Miller, J., & Bauer, D. W. (1981). Irrelevant differences in the “same”—“

different”task. Journal of Experimental Psychology: Human Perception

and Performance,7(1), 196–207. https://doi.org/10.1037//0096-1523.7

.1.196

Morey, R. D. (2008). Confidence intervals from normalised data: A correc-

tion to Cousineau (2005). Tutorials in Quantitative Methods for Psy-

chology,4(2), 61–64. https://doi.org/10.20982/tqmp.04.2.p061

Nickerson, R. S. (1965). Response times for “Same”“Different”judg-

ments. Perceptual and Motor Skills,20,15–18. https://doi.org/10.2466/

pms.1965.20.1.15

Nickerson, R. S. (1967a). Categorization time with categories defined by

disjunctions and conjunctions of stimulus attributes. Journal of Experi-

mental Psychology,73(2), 211–219. https://doi.org/10.1037/h0021270

Nickerson, R. S. (1967b). “Same”-“different”response times with multi-at-

tribute stimulus differences. Perceptual and Motor Skills,24(2), 543–554.

https://doi.org/10.2466/pms.1967.24.2.543

Nickerson, R. S. (1968). Note on “Same”-“Different”response times. Per-

ceptual and Motor Skills,27(2), 565–566. https://doi.org/10.2466/pms

.1968.27.2.565

Nickerson, R. S. (1978). On the time it takes to tell things apart. In J.

Requin (Ed.), Attention and performance VII (pp. 77–88). Erlbaum.

Nickerson, R. S., & Pew, R. W. (1973). Visual pattern matching: An inves-

tigation of some effects of decision task, auditory codability, and spatial

correspondence. Journal of Experimental Psychology,98(1), 36–43.

https://doi.org/10.1037/h0034300

Petrov, A. A. (2009). Symmetry-based methodology for decision-rule

identification in same—Different experiments. Psychonomic Bulletin &

Review,16(6), 1011–1025. https://doi.org/10.3758/PBR.16.6.1011

Posner, M. I., & Boies, S. J. (1971). Components of attention. Psychologi-

cal Review,78(5), 391–408. https://doi.org/10.1037/h0031333

Posner, M. I., & Mitchell, R. F. (1967). Chronometric analysis of classifi-

cation. Psychological Review,74(5), 392–409. https://doi.org/10.1037/

h0024913

Potter, K. W., Donkin, C., & Huber, D. E. (2018). The elimination of posi-

tive priming with increasing prime duration reflects a transition from

perceptual fluency to disfluency rather than bias against primed words.

Cognitive Psychology,101,1–28. https://doi.org/10.1016/j.cogpsych

.2017.11.004

Proctor, R. W. (1981). A unified theory for matching-task phenomena.

Psychological Review,88(4), 291–326. https://doi.org/10.1037/0033

-295X.88.4.291

Proctor, R. W. (1986). Response bias, criteria settings, and the fast-same

phenomenon: A reply to Ratcliff. Psychological Review,93(4),

473–477. https://doi.org/10.1037/0033-295X.93.4.473

Proctor, R. W., & Rao, K. V. (1982). On the “misguided”use of reaction-

time differences: A discussion of Ratcliff and Hacker (1981). Perception

& Psychophysics,31(6), 601–604. https://doi.org/10.3758/bf03204200

Proctor, R. W., & Rao, K. V. (1983a). Evidence that the same-different

disparity in letter matching is not attributable to response bias. Percep-

tion & Psychophysics,34(1), 72–76. https://doi.org/10.3758/bf03205898

Proctor, R. W., & Rao, K. V. (1983b). Reinstating the original principles

of Proctor’s unified theory for matching-task phenomena: An evaluation

of Krueger and Shapiro’s reformulation. Psychological Review,90(1),

21–37. https://doi.org/10.1037/0033-295X.90.1.21

Proctor, R. W., Rao, K. V., & Hurst, P. W. (1984). An examination of

response bias in multiletter matching. Perception & Psychophysics,

35(5), 464–476. https://doi.org/10.3758/bf03203923

Ratcliff, R. (1978). A theory of memory retrieval. Psychological Review,

85(2), 59–108. https://doi.org/10.1037/0033-295X.85.2.59

Ratcliff, R. (1985). Theoretical interpretations of the speed and accuracy

of positive and negative responses. Psychological Review,92(2),

212–225. https://doi.org/10.1037/0033-295X.92.2.212

Ratcliff, R., & Hacker, M. J. (1981). Speed and accuracy of same and dif-

ferent responses in perceptual matching. Perception & Psychophysics,

30(3), 303–307. https://doi.org/10.3758/bf03214286

Ratcliff, R., & Hacker, M. J. (1983). On the misguided use of reaction-

time differences: A reply to Proctor & Rao (1982). Perception & Psy-

chophysics,31(6), 603–604. https://doi.org/10.3758/BF03204201

Sekuler, R. W., & Abrams, M. (1968). Visual sameness: A choice time

analysis of pattern recognition processes. Journal of Experimental Psy-

chology,77(2), 232–238. https://doi.org/10.1037/h0025741

Silverman, W. P., & Goldberg, S. L. (1975). Further confirmation of same

vs. different processing differences. Perception & Psychophysics,17(2),

189–193. https://doi.org/10.3758/BF03203884

Snodgrass, J. G. (1972). Matching patterns vs matching digits: The effect

of memory dependence and complexity on “Same”-“Different”reaction

times. Perception & Psychophysics,11(5), 341–349. https://doi.org/10

.3758/BF03206264

St. James, J. D., & Eriksen, C. W. (1991). Response competition produces

a“fast same effect”in same-different judgments. In G. R. Lockhead,

J. R. Pomerantz (Eds.), The perception of structure: Essays in honor of

Wendell R. Garner (pp. 157–168). American Psychological Association.

https://doi.org/10.1037/10101-009

Sternberg, S. (1998). Inferring mental operations from reaction-time data:

how we compare objects. In D. N. Osherson, D. Scarborough, & S.

RESIDUAL ACTIVATION PRIMING 537

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Sternberg (Eds.), An invitation to cognitive science, Vol. 4: Methods,

models, and conceptual issues (pp. 365–454). MIT Press.

Swan, G., & Wyble, B. (2014). The binding pool: A model of shared neu-

ral resources for distinct items in visual working memory. Attention,

Perception, & Psychophysics,76(7), 2136–2157. https://doi.org/10

.3758/s13414-014-0633-3

Taylor, D. A. (1976a). Effect of identity in the multiletter matching

task. Journal of Experimental Psychology: Human Perception and

Performance,2(3), 417–428. https://doi.org/10.1037/0096-1523.2.3

.417

Taylor, D. A. (1976b). Holistic and analytic processes in the comparison

of letters. Perception & Psychophysics,20(3), 187–190. https://doi.org/

10.3758/BF03198599

Taylor, R. L. (1969). Comparison of short-term memory and visual sensory

analysis as sources of information. Journal of Experimental Psychology,

81(3), 515–522. https://doi.org/10.1037/h0027894

Townsend, J. T., & Ashby, F. G. (1983). The stochastic modeling of ele-

mentary psychological processes. Cambridge University Press.

Townsend, J. T., & Nozawa, G. (1995). Spatio-temporal properties of ele-

mentary perception an investigation of parallel, serial, and coactive theo-

ries. Journal of Mathematical Psychology,39(4), 321–359. https://doi

.org/10.1006/jmps.1995.1033

Treisman, A. M. (1985). Pre-attentive processing in vision. Computer

Vision Graphics and Image Processing,31(2), 156–177. https://doi.org/

10.1016/S0734-189X(85)80004-9

Tversky, B. (1969). Pictorial and verbal encoding in a short-term memory

task. Perception & Psychophysics,6(4), 225–233. https://doi.org/10

.3758/BF03207022

van Opstal, F., & Verguts, T. (2011). The origins of the numerical distance

effect: The same–different task. Journal of Cognitive Psychology,23(1),

112–120. https://doi.org/10.1080/20445911.2011.466796

Wagenmakers, E. J., van der Maas, H. L. J., & Grasman, R. P. P. P.

(2007). An EZ-diffusion model for response time and accuracy. Psycho-

nomic Bulletin & Review,14(1), 3–22. https://doi.org/10.3758/bf03194023

Walker, J. A., & Cousineau, D. (2019). Into the mind’s eye: Exploring the

fast–same effect in the Same–Different task. The American Journal of

Psychology,132(4), 421–437. https://doi.org/10.5406/amerjpsyc.132.4

.0421

Well, A. D., Pollatsek, A., & Schindler, R. M. (1975). Facilitation of both

same and different judgments of letter strings by familiarity of letter

sequence. Perception & Psychophysics,17(5), 511–520. https://doi.org/

10.3758/BF03203303

Wolfe, J. M. (1994). Guided Search 2.0: A revised model of visual search.

Psychonomic Bulletin & Review,1(2), 202–238. https://doi.org/10.3758/

BF03200774

Appendix A

Higher-Moment Statistics of the Four Experiments

In this appendix, we report the higher-moment statistics for all

four variants of the same-different task found in the main text.

The main goal of these analyses is to determine if a qualitative

change exists in distributional properties across priming condi-

tions. If there is such a change, an important shift in processing

mode between experimental manipulations could have occurred.

On the other hand, if priming only affects the processing latency

of “same”decisions, there should be no changes to the underly-

ing decision-making mechanism of the task and thus no noticea-

ble qualitative differences for higher-moment statistics across

experiments and priming conditions.

In the following, we examined mean standard deviation and

mean skewness of all 14 stimulus composition conditions

within each priming condition. Similar to the mean RT analy-

ses reported in the main text, data were aggregated by partici-

pants and averaged for all stimulus composition conditions.

The error bars denote the within-subject 95% conﬁdence inter-

val (CI) using the appropriate CI estimator (the standard devia-

tion CI estimator for standard deviations and the skewness CI

estimator for skew; Harding et al., 2014). Note that CIs for

standard deviation are asymmetrical as the critical values are

taken from the vdistribution, an asymmetrical distribution.

Conversely, the CIs for skewness are all identical in size

because its estimator’s calculation depends only on the sample

size (number of participants per experiment) and are therefore

identical across stimulus composition conditions.

Experiment 1: Case Manipulation

Results regarding Experiment 1 are shown in Figure A1.As

seen in the pattern of results for both standard deviation and

skew are similar across priming conditions. To verify whether

there was a signiﬁcant difference across these priming condi-

tions, we performed a repeated-measures 14 32 ANOVA on

the mean standard deviation measures of all “same”and “differ-

ent”trials with a factor for the 14 stimulus composition condi-

tions and a factor for the two priming conditions. We found a

nonsigniﬁcant effect for priming, F(1, 18) = 3.755, p=.068,h

=.173,asigniﬁcant effect of stimulus composition, F(13, 221) =

17.471, p=.001,h

= .134, and a nonsigniﬁcant interaction

between both factors, F(13, 234) = 1.088, p=.370,h

=.057.

Another 14 32 ANOVA was performed on the mean skewness

measures with identical factors to the previous ANOVA. We

found a nonsigniﬁcant effect of priming, F(1, 18) = 1.112, p=

.306, h

=.058,asigniﬁcant effect of stimulus composition, F

(13, 234) = 17.471, p,.001, h

= .336, and a nonsigniﬁcant

interaction between factors, F(13, 234) = 1.477, p=.139,h

.074. We can thus conclude that cancelling priming did not cause

noticeable changes to the RT distributions and thus did not affect

the underlying decision-making process in this experiment.

Experiment 2: Font Manipulation

Mean measures of standard deviation and skewness for both

priming conditions of Experiment 2 are shown in Figure A2.

Results regarding this experiment yield little supplemental infor-

(Appendices continue)

538 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

mation beyond the results of Experiment 1. For mean standard

deviation measures, a 14 32 repeated-measures ANOVA with a

factor for the stimulus composition conditions and one for the pri-

ming conditions revealed a nonsigniﬁcant effect of priming, F(1,

19) = .062, p= .807, h

= .003, a signiﬁcant effect of stimulus

composition, F(13, 247) = 2.920, p= .001, h

= .133, and a non-

signiﬁcant interaction between factors, F(13, 247) = .550, p=

.891, h

= .028. For mean skewness measures, a 14 32repeated-

measures ANOVA revealed a nonsigniﬁcant effect of priming, F

(1, 19) = .621, p= .440, h

= .032, a signiﬁcant effect of stimulus

composition, F(13, 247) = 6.162, p,.001, h

= .245, and a non-

signiﬁcant interaction between factors, F(13, 247) = .732, p=

.730, h

= .037. Again, we can conclude that the lack of a signiﬁ-

cant difference between priming conditions and the lack of interac-

tion indicates that it is likely the same underlying decision-making

process that is present for both priming conditions.

Experiment 3: Visual Versus Auditory Stimuli

Results for Experiment 3 are presented in Figure A3.As

shown, results from visual stimuli do not differ from the control

conditions seen so far. Regarding auditory stimuli, the standard

deviation follows an upward trend as the string’slengthgrows.

This abnormality could also be due to interference from the serial

nature of S

(the presentation of another stimulus hampers the

recall of previously presented stimuli) and decay (stimuli that

have been held longer in memory may be harder to recall) of the

mental representation over time. As for skewness, there are no

visual differences between priming conditions worth noting.

A repeated-measures 14 32 ANOVA on mean standard

deviation measures with a factor for the stimulus composition

conditions and another factor for the priming conditions

showed that there is a signiﬁcant effect of priming, F(1, 18) =

31.781, p,.001, h

= .638, a signiﬁcant effect of stimulus

composition, F(13, 234) = 4.387, p,.001, h

= .282, and a

signiﬁcant interaction between factors, F(13, 234) = 5.330, p,

.001, h

= .228. This interaction is due to the major changes

in standard deviation measures between priming conditions

(mean standard deviation for visual stimuli is 199.33 ms,

whereas it is 270.89 ms for auditory trials; additionally, visual

trials have a relatively ﬂat slope compared to the deﬁnite posi-

tive slope of their auditory counterparts). Another 14 32

repeated-measures ANOVA for mean skewness measures

showed that there is a signiﬁcant effect of priming, F(1, 18) =

Figure A1

Mean Standard Deviation (Top Panels) and Mean Skewness (Bottom Panels) Results for

Experiment 1

Note. The left panels show the results for matching stimuli, and mismatching stimuli are in the right panels.

Error bars denote the correlation- and difference-adjusted 95% conﬁdence interval of the mean (Baguley, 2012)

constructed with the proper estimator (Harding et al., 2014). D = differences. See the online article for the color

version of this ﬁgure.

(Appendices continue)

RESIDUAL ACTIVATION PRIMING 539

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

38.475, p,.001, h

= .681, a signiﬁcant effect of stimulus

composition, F(13, 234) = 7.071, p,.001, h

= .196, and no

signiﬁcant interaction between factors, F(13, 234) = .878, p=

.577, h

= .046.

The signiﬁcant priming effect in the ANOVAs of both

higher-moment statistics goes against our initial hypothesis.

However, due to the change in encoding modality (forcibly se-

rial for auditory trials vs. possibly all at once for visual trials),

these signiﬁcant effects do not take away from our main ﬁnding

that a complete visual match is required for a fast-same effect

to occur.

Experiment 4: Long-Term Memory Associations

The results for Experiment 4 are presented in Figure A4.As

seen, results resemble those of Experiments 1 and 2, and both

mean standard deviation and mean skewness seem relatively

stable across priming conditions. A 14 32 repeated-measures

ANOVA performed on mean standard deviation measures with

a factor for the stimulus composition conditions and another

factor for the priming conditions yielded a signiﬁcant effect of

priming, F(1, 18) = 13.529, p= .002, h

= .429, a signiﬁcant

effect of stimulus composition, F(13, 221) = 2.384, p= .005,

= .117, and a nonsigniﬁcant interaction between factors, F

(13, 234) = 1.350, p= .185, h

= .070. Another 14 32

repeated-measures ANOVA was performed on mean skewness

measures, which resulted in a signiﬁcant effect of priming, F(1,

18) = 5.383, p= .032, h

= .230, a signiﬁcant effect of stimulus

composition, F(13, 234) = 4.547, p,.001, h

= .202, and a

nonsigniﬁcant interaction between factors, F(13, 234) = 1.440,

p= .142, h

= .074.

Again, these signiﬁcant differences between priming condi-

tions for both higher-moment statistics are incompatible with our

original hypothesis. Nevertheless, the changes in higher-moment

statistics are likely due to the change in task demands (from an

initial comparison task to a LTM retrieval þcomparison task).

Our take-home message remains the same: A complete visual

match is necessary for the fast-same phenomenon to occur.

Discussion

The measures of spread and skew are conclusive in their

inconclusiveness. There are no noticeable qualitative changes

between the four experiments other than the mean standard

deviation measures of auditory stimuli (but this change could be

due to the unusual, serial presentation of S

). This suggests that

Figure A2

Mean Standard Deviation (Top Panels) and Mean Skewness (Bottom Panels) Results for Experiment 2

Note. The left panels show the results for matching stimuli, and mismatching stimuli are in the right panels. Error

bars denote the correlation- and difference-adjusted 95% conﬁdence interval (Baguley, 2012) constructed with the

proper estimator (Harding et al., 2014). D = differences. See the online article for the color version of this ﬁgure.

(Appendices continue)

540 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

the way in which participants process stimuli and compare ele-

ments is relatively rigid across priming conditions. All experi-

ments offer the same, relatively ﬂat slope across string length

for fast-same and attenuated-same conditions alike. A change in

processing mode across priming conditions would more than

likely have resulted in some sort of modiﬁcation to either de-

scriptive statistic, especially if we consider Bamber’s (1969)

original hypothesis that “sameness”can be assessed by a fast-

parallel process, whereas “differentness”must be assessed by a

serial process (as proposed with his identity reporter vs. serial

self-termination model). This lack of change suggests that both

“same”and “different”decisions likely operate with identical,

or at least quite similar, underlying processes.

While there are no clear trends in either analysis (along

with a lack of consistency among the ANOVAs performed in

each experiment), these results still offer a glimpse into the

cognitive architecture at play for both decisions. A decrease in

skewness typically indicates that the RT distribution tends to-

ward the normal distribution as stimuli sizes increase due to the

central limit theorem, a theorem strongly associated to serial

architectures. However, the fact that mean skewness measures

remain fairly constant across string length favors parallel proc-

essing. Likewise, for standard deviation, a lack of visible

changes as string lengths increase suggests a parallel process as

well. Standard deviation would tend to increase when the num-

ber of comparisons is increased (as is the case for serial archi-

tectures) when processing is not exhaustive.

Consequently, (a) processing mode seems constant across

priming conditions, and (b) although identifying the processing

architecture is not the focus of this article, it seems plausible that

the comparison process follows parallel architecture (but see

Harding et al., 2021). Note that we do not make strong claims

regarding architecture diagnosis here as we (mostly) observed

null effects. Without an appropriate power analysis, we have no

idea how strongly these claims can be afﬁrmed. Further, it is

known that skew analyses generally lack statistical power.

Figure A3

Mean Standard Deviation (Top Panels) and Mean Skewness (Bottom Panels) Results for Experiment 3

Note. The left panels show the results for visually presented trials, and auditory trials are in the right panels.

Error bars denote the correlation- and difference-adjusted 95% conﬁdence interval (Baguley, 2012) constructed

with the proper estimator (Harding et al., 2014). D = differences. See the online article for the color version of

this ﬁgure.

(Appendices continue)

RESIDUAL ACTIVATION PRIMING 541

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Appendix B

EZ Parameter Predictions of Each Same-Different Task Variant

In this appendix, we analyze the results of all four experiments

using the EZ diffusion model (Wagenmakers et al., 2007). We

wish to determine whether the fast-same effect, and its attenua-

tion, from all four experiments are predicted by a change in

threshold, accumulation rate, and/or residual time (that is,

encoding and motor time). The EZ model estimates these pa-

rameters (noted vfor accumulation, afor threshold, and T

for

residual time), requiring only the accuracy rate, the mean RT,

and the RT variance from the targeted conditions to do so.

Parameter estimations were performed using EZ4Math (T.

Groulx et al., 2020), an EZ module for the open-access

Wolfram Language and conﬁrmed with the ezDiffusion pack-

age for R (Version 0.1.0; Barth, 2019). The analyses were per-

formed on an individual basis and averaged across participants.

When a participant had perfect accuracy for any given stimulus

composition condition, it was replaced with 99.9% because EZ

cannot make estimates in these situations. The scaling parame-

ter, s, was set to the customary .1.

The averaged estimates across participants of all three pa-

rameters for all four experiments are given in Table B1. These

estimates are broken down by priming conditions (control vs.

experimental conditions; see main text for their notation) as

well as by the 14 stimulus composition conditions. Each esti-

mate is accompanied by the standard error of the mean (in

parentheses).

As is seen in Table B1, there is very little variation for both

aand vparameters between each stimulus composition condi-

tion across priming conditions for each of the four experiments.

In fact, the only notable change between these experimental

conditions stems from the T

parameter. This would therefore

signify that EZ predicts the changes in RT between “same”and

“different”decisions almost entirely as changes in nondecision

time for all four experiments. Therefore, EZ predicts similar

RT distribution shapes across each experiment’s 28 total exper-

imental conditions (14 stimulus composition conditions and

two priming conditions for each), much like what was found in

Figure A4

Mean Standard Deviation (Top Panels) and Mean Skewness (Bottom Panels) Results for Experiment 4

Note. The left panels show the results for S

-shown trials, and S

-absent trials are in the right panels. Error bars

denote the correlation- and difference-adjusted 95% conﬁdence interval (Baguley, 2012) constructed with the

proper estimator (Harding et al., 2014). D = differences. See the online article for the color version of this ﬁgure.

(Appendices continue)

542 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table B1

EZ Parameter Estimates for v, a, and T

of All 14 Stimulus Composition Conditions for Both Priming Conditions for All Four Experiments

String length Same 1D 2D 3D 4D

Experiment 1 matching (vparameter estimates)

1 0.000304 (0.000024) 0.000285 (0.000020)

2 0.000273 (0.000015) 0.000295 (0.000020) 0.000325 (0.000029)

3 0.000289 (0.000019) 0.000218 (0.000017) 0.000325 (0.000026) 0.000388 (0.000031)

4 0.000273 (0.000017) 0.000144 (0.000020) 0.000365 (0.000020) 0.000407 (0.000033) 0.000358 (0.000032)

Experiment 1 mismatching (vparameter estimates)

1 0.000281 (0.000015) 0.000294 (0.000016)

2 0.000289 (0.000019) 0.000258 (0.000017) 0.000319 (0.000026)

3 0.000275 (0.000018) 0.000186 (0.000013) 0.000298 (0.000019) 0.000383 (0.000029)

4 0.000232 (0.000012) 0.000111 (0.000021) 0.000274 (0.000020) 0.000299 (0.000023) 0.000370 (0.000032)

Experiment 2 matching (vparameter estimates)

1 0.000294 (0.000022) 0.000293 (0.000021)

2 0.000277 (0.000021) 0.000267 (0.000019) 0.000318 (0.000030)

3 0.000264 (0.000017) 0.000226 (0.000020) 0.000329 (0.000022) 0.000340 (0.000020)

4 0.000271 (0.000015) 0.000127 (0.000021) 0.000261 (0.000029) 0.000354 (0.000035) 0.000367 (0.000027)

Experiment 2 mismatching (vparameter estimates)

1 0.000271 (0.000015) 0.000292 (0.000022)

2 0.000268 (0.000015) 0.000271 (0.000022) 0.000320 (0.000020)

3 0.000257 (0.000011) 0.000212 (0.000019) 0.000309 (0.000027) 0.000348 (0.000028)

4 0.000236 (0.000014) 0.000150 (0.000016) 0.000280 (0.000027) 0.000354 (0.000034) 0.000353 (0.000026)

Experiment 3 matching (vparameter estimates)

1 0.000265 (0.000018) 0.000233 (0.000014)

2 0.000262 (0.000020) 0.000253 (0.000026) 0.000303 (0.000030)

3 0.000256 (0.000016) 0.000235 (0.000024) 0.000309 (0.000026) 0.000284 (0.000029)

4 0.000273 (0.000027) 0.000136 (0.000029) 0.000283 (0.000023) 0.000334 (0.000034) 0.000279 (0.000030)

Experiment 3 mismatching (vparameter estimates)

1 0.000249 (0.000021) 0.000254 (0.000020)

2 0.000214 (0.000013) 0.000209 (0.000012) 0.000300 (0.000026)

3 0.000191 (0.000014) 0.000150 (0.000017) 0.000261 (0.000023) 0.000307 (0.000025)

4 0.000163 (0.000008) 0.000095 (0.000013) 0.000199 (0.000018) 0.000292 (0.000023) 0.000303 (0.000028)

Experiment 4 matching (vparameter estimates)

1 0.000295 (0.000022) 0.000304 (0.000034)

2 0.000303 (0.000027) 0.000303 (0.000032) 0.000311 (0.000039)

3 0.000285 (0.000020) 0.000267 (0.000027) 0.000355 (0.000037) 0.000383 (0.000040)

4 0.000301 (0.000030) 0.000218 (0.000030) 0.000329 (0.000036) 0.000370 (0.000033) 0.000366 (0.000033)

Experiment 4 mismatching (vparameter estimates)

1 0.000248 (0.000023) 0.000244 (0.000022)

2 0.000238 (0.000022) 0.000224 (0.000018) 0.000279 (0.000022)

3 0.000240 (0.000021) 0.000220 (0.000028) 0.000308 (0.000033) 0.000312 (0.000028)

4 0.000240 (0.000019) 0.000182 (0.000028) 0.000284 (0.000020) 0.000332 (0.000039) 0.000372 (0.000036)

Experiment 1 matching (aparameter estimates)

1 0.161809 (0.011620) 0.154650 (0.012055)

2 0.130294 (0.006928) 0.147347 (0.014479) 0.151959 (0.014196)

3 0.132666 (0.006984) 0.127112 (0.012622) 0.161958 (0.015833) 0.169116 (0.013010)

4 0.138222 (0.006161) 0.116488 (0.011556) 0.169049 (0.011050) 0.145258 (0.011129) 0.155521 (0.015242)

Experiment 1 mismatching (aparameter estimates)

1 0.131229 (0.009498) 0.163767 (0.014402)

2 0.146571 (0.009400) 0.139100 (0.009049) 0.164155 (0.011354)

3 0.125023 (0.009335) 0.130202 (0.012212) 0.167540 (0.015217) 0.192665 (0.016346)

4 0.123955 (0.005777) 0.117981 (0.009750) 0.146064 (0.012743) 0.154431 (0.014343) 0.153672 (0.010059)

Experiment 2 matching (aparameter estimates)

1 0.167842 (0.011932) 0.159176 (0.014057)

2 0.137341 (0.008965) 0.165312 (0.017096) 0.167505 (0.017369)

3 0.151406 (0.014314) 0.148792 (0.017365) 0.179700 (0.014405) 0.169562 (0.014752)

4 0.154846 (0.014549) 0.123362 (0.013522) 0.161434 (0.017010) 0.177598 (0.012167) 0.161175 (0.011663)

Experiment 2 mismatching (aparameter estimates)

1 0.159846 (0.010447) 0.162297 (0.010395)

2 0.144659 (0.012710) 0.155690 (0.014010) 0.156914 (0.014911)

3 0.146726 (0.014793) 0.160618 (0.018751) 0.180435 (0.012033) 0.178461 (0.012921)

4 0.135995 (0.008245) 0.117209 (0.010122) 0.157714 (0.011410) 0.169997 (0.018063) 0.177675 (0.015605)

(Appendices continue)

RESIDUAL ACTIVATION PRIMING 543

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table B1 (continued)

String length Same 1D 2D 3D 4D

Experiment 3 matching (aparameter estimates)

1 0.166230 (0.011959) 0.198363 (0.015133)

2 0.138300 (0.009828) 0.152438 (0.010781) 0.167420 (0.011189)

3 0.160245 (0.011362) 0.163459 (0.015526) 0.174445 (0.017951) 0.157970 (0.015062)

4 0.151800 (0.009799) 0.122790 (0.011047) 0.165276 (0.017818) 0.160374 (0.014735) 0.179977 (0.016044)

Experiment 3 mismatching (aparameter estimates)

1 0.143929 (0.009028) 0.173228 (0.012939)

2 0.164154 (0.012098) 0.179083 (0.016372) 0.211054 (0.016745)

3 0.162987 (0.010853) 0.157476 (0.011339) 0.203751 (0.016965) 0.220639 (0.020227)

4 0.163516 (0.007336) 0.162433 (0.016578) 0.251168 (0.020516) 0.233019 (0.017941) 0.205496 (0.015981)

Experiment 4 matching (aparameter estimates)

1 0.164148 (0.018492) 0.157818 (0.016027)

2 0.155468 (0.014223) 0.191840 (0.018535) 0.186269 (0.013435)

3 0.159341 (0.015219) 0.164493 (0.018806) 0.178012 (0.013558) 0.172228 (0.016477)

4 0.170800 (0.016468) 0.139989 (0.018204) 0.192690 (0.021314) 0.159388 (0.015099) 0.169770 (0.018210)

Experiment 4 mismatching (aparameter estimates)

1 0.159584 (0.014801) 0.183320 (0.017995)

2 0.159811 (0.015204) 0.175758 (0.019477) 0.191777 (0.019603)

3 0.169136 (0.019375) 0.142846 (0.010930) 0.202370 (0.018441) 0.211538 (0.017232)

4 0.160879 (0.016956) 0.155540 (0.020563) 0.210517 (0.019112) 0.197267 (0.017232) 0.170871 (0.017756)

Experiment 1 matching (T

parameter estimates)

1 201 (19) 251 (23)

2 260 (17) 307 (32) 275 (25)

3 276 (8) 365 (19) 275 (24) 271 (23)

4 274 (12) 382 (23) 325 (14) 332 (14) 276 (24)

Experiment 1 mismatching (T

parameter estimates)

1 266 (15) 244 (25)

2 274 (16) 323 (17) 251 (19)

3 323 (13) 356 (22) 284 (25) 219 (35)

4 336 (18) 386 (15) 323 (18) 288 (18) 288 (19)

Experiment 2 matching (T

parameter estimates)

1 206 (20) 288 (24)

2 271 (18) 321 (27) 277 (32)

3 267 (27) 366 (20) 305 (26) 305 (30)

4 289 (29) 409 (41) 326 (34) 294 (33) 329 (23)

Experiment 2 mismatching (T

parameter estimates)

1 221 (16) 284 (17)

2 272 (23) 324 (30) 306 (22)

3 300 (26) 331 (34) 291 (29) 279 (16)

4 345 (25) 416 (25) 346 (22) 303 (38) 286 (35)

Experiment 3 matching (T

parameter estimates)

1 203 (24) 199 (29)

2 240 (17) 323 (15) 270 (22)

3 252 (23) 347 (32) 271 (29) 283 (31)

4 283 (18) 384 (31) 311 (37) 313 (20) 253 (37)

Experiment 3 mismatching (T

parameter estimates)

1 247 (14) 288 (24)

2 276 (22) 345 (33) 253 (28)

3 392 (27) 499 (29) 338 (31) 282 (36)

4 530 (37) 579 (46) 298 (40) 321 (35) 362 (40)

Experiment 4 matching (T

parameter estimates)

1 219 (24) 301 (25)

2 246 (20) 251 (41) 244 (34)

3 253 (13) 313 (38) 310 (19) 287 (18)

4 252 (21) 395 (36) 280 (38) 343 (29) 297 (37)

Experiment 4 mismatching (T

parameter estimates)

1 270 (14) 283 (26)

2 326 (23) 341 (27) 294 (31)

3 331 (20) 440 (32) 294 (30) 256 (27)

4 343 (22) 409 (29) 327 (41) 292 (36) 335 (22)

Note. D = differences; v = accumulation rate; a = threshold; T

= residual time. All results represent the mean value across participants along with the

standard error in parentheses.

(Appendices continue)

544 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Appendix A (the decision parameters of the model, aand v, are

the components that predict changes in RT distribution spread

and skew). This ﬁnding is in line with the priming hypothesis

posited in the main body of the text—if the decision can begin

more quickly (as hypothesized by residual activation; see the

nROUSE model of Huber, 2008), then there is less encoding

time necessary (encoding time being one of the components

that contribute to the T

parameter). Furthermore, these

ﬁndings corroborate the results of Potter et al. (2018) in a simi-

lar task. In their work, it was also discovered that overall RTs

can be predicted by nondecision time.

To further emphasize these results, the averaged estimates

(along with their 95% CI) of all four experiments for each of

the three parameters are given in Figure B1 for both priming

conditions. These estimates are also broken down into overall

“same”and overall “different”conditions (the individual

Figure B1

EZ Parameter Estimates for v, a, and T

(Individual Columns) Subdivided by Experiment (Individual Rows)

Note. All results represent the mean value across participants along with their 95% conﬁdence interval. Points representing the means of “same”deci-

sions are composed of four stimulus composition conditions, whereas points representing the means of “different”decisions are composed of 10 stimulus

composition conditions. See the online article for the color version of this ﬁgure.

(Appendices continue)

RESIDUAL ACTIVATION PRIMING 545

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

lines). For estimates regarding “same”conditions, the overall

average is taken from the four stimulus composition condi-

tions, whereas for “different,”it is taken from the 10 stimulus

composition conditions. In this ﬁgure, each column denotes

estimates of a given parameter, and every row denotes results

from that experiment (the number referring to the experiment

number found in the main text).

We can see that T

estimates are smaller for matching

“same”responses than for matching “different”responses by

about 50–60 ms; this is about the magnitude of the empirical

fast-same effect found in the main body of the text (45 ms in

Experiment 1 and 60 ms in Experiment 2). When comparing

estimates between mismatching “same”and mismatching

“different”conditions, we can see that they are much closer in

value, mirroring the smaller Hedges’gfound in the main text

when priming was annulled. Nevertheless, while empirical

“same”RTs are always faster than empirical “different”RTs in

the experimental conditions, T

estimates only mirror this

trend in Experiments 1 and 2, whereas in Experiments 3 and 4,

is higher for “same”than it is for “different”(this effect is

very pronounced in the estimates of Experiment 3, whereas

they are more subtle in the estimates of Experiment 4).

Considering that EZ seems to be sensitive to changes in RT dis-

tribution shapes, and that, as noted in Appendix A, there is

much more variability between higher-moment statistics meas-

ures in the two last experiments, it is not surprising that EZ pre-

dicts a different trend for these experiments. Finally, it seems

like the predicted T

for “different”responses is near identical

across priming conditions for all four experiments. Regarding

threshold estimates, they seem stable between priming condi-

tions. There might be a trend for “same”thresholds to be

smaller than “different”thresholds throughout all four experi-

ments, but this difference is only particularly noteworthy in

Experiment 1’s mismatching condition. This overall trend

might reﬂect a small bias for “same”responses and be reﬂec-

tive of the smaller intercepts for “same”conditions, as revealed

by the main text’s slope analyses. Finally, for accumulation

rate, matching “same”and matching “different”estimates have

similar values, and there is little to no change between “differ-

ent”estimates across priming conditions for all four experi-

ments. However, it seems as though EZ predicts a lower

accrual rate for mismatching “same”conditions when com-

pared to the other three decision types. Thus, these EZ parame-

ter estimates lead to the conclusion that the change in latency

between priming conditions for “same”decisions is barely

modulated by a change in the threshold and is weakly affected

by the accumulation rate (slower speed for mismatching

“same”trials). The most important predictor of mean RT is

indeed T

. In fact, when comparing the 28 empirical mean RTs

and the 28 estimates of T

for each experiment, we observe an

average correlation of .84 (ranging from .74 for experiment 4

to .91 for Experiment 2). For a,weﬁnd an average correlation

of .12 between the 28 empirical mean RTs and the 28 esti-

mates (ranging from .47 for Experiment 1 to .31 for

Experiment 3); for v,weﬁnd an average correlation of .68

between the empirical RTs and EZ’s estimates (ranging from

.64 for Experiment 4 to .74 for Experiment 1). This result

could not be caused by estimation biases in the parameter

search as EZ only relies on an algebraic relation linking mean,

variance, and percent correct to create the parameter estimates.

The model predictions found here might ﬁt with the fast-track

hypothesis: Repeated stimulus in the matching “same”condi-

tions get a free ride along the processing pathway—the similar-

ity between S

and S

enables stimuli to be processed just

before inhibitory processes can be built up, resulting in a clear

passage up to the decision-making module. It remains to be

investigated how/why this explanation would not inﬂuence

processing rate and thresholds more.

While this appendix is not a fully-ﬂedged modeling attempt

of same-different task results, similar results were also reported

in Groulx et al. (2020) with another same-different task variant.

Goulet & Cousineau (2019) tried more advanced sampling

models on another same-different task variant resulting in simi-

lar results. These many modeling attempts indicate that ﬁtting

same-different task results will prove to be challenging if we

look toward explaining the changes in RT with aspects of the

decision mechanism other than nondecision times. Existing

models have difﬁculties attributing the fast-same effect to one

of the hypothesized constructs (rates, thresholds, base response

times), and modeling performance in this task will be a central

aspect of our future endeavors.

Received January 3, 2020

Revision received May 27, 2021

Accepted July 8, 2021 n

546 HARDING AND COUSINEAU

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

A preview of this full-text is provided by American Psychological Association.

Learn more

Content available from Journal of Experimental Psychology Learning Memory and Cognition

This content is subject to copyright. Terms and conditions apply.

The Perception of Similarity, Difference and Opposition

Article

Full-text available

Aug 2023

After considering the pervasiveness of same/different relationships in Psychology and the experimental evidence of their perceptual foundation in Psychophysics and Infant and Comparative Psychology, this paper develops its main argument. Similarity and diversity do not complete the panorama since opposition constitutes a third relationship which is distinct from the other two. There is evidence of this in the previous literature investigating the perceptual basis of opposition and in the results of the two new studies presented in this paper. In these studies, the participants were asked to indicate to what extent pairs of simple bi-dimensional figures appeared to be similar, different or opposite to each other. A rating task was used in Study 1 and a pair comparison task was used in Study 2. Three main results consistently emerged: Firstly, opposition is distinct from similarity and difference which, conversely, are in a strictly inverse relationship. Secondly, opposition is specifically linked to something which points in an allocentrically opposite direction. Thirdly, alterations to the shape of an object are usually associated with the perception of diversity rather than opposition. The implications of a shift from a dyadic (same/different) to a triadic (similar/different/opposite) paradigm are discussed in the final section.

Analyses of Response Time Data in the Same–Different Task

Article

Full-text available

Apr 2023

The Same-Different task presents two stimuli in close succession and participants must indicate whether they are completely identical or if there are any attributes that differ. While the task is simple, its results have proven difficult to explain. Notably, response times are characterized by a fast-same effect whereby Same responses are faster than Different responses even though identical stimuli should be exhaustively processed to be accurate. Herein, we examine a little more than a quarter million response times (N = 255,744) obtained from 327 participants who participated in one of 14 variants of the task involving minor changes in the stimuli or their durations. We performed distribution fitting and analyzed estimated parameters stemming from the ex-Gaussian, lognormal, and Weibull distributions to infer the cognitive processing characteristics underlying this task. The results exclude serial processing of the stimuli and do not support dual-route processing. The fast-same effect appears only through a shift of the entire response time distributions, a feature impossible to detect solely with mean response time analyses. An attention-modulated process driven by entropy may be the most adequate model of the fast-same effect. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

“Same”–“Different” Judgments: A Review of Current Controversies in Perceptual Comparisons

Article

Full-text available

Nov 1985

Bart Farell

A commonly used experimental paradigm for studying human information processing asks subjects to classify two stimuli with respect to their similarities and differences. The “same”-“different” judgment task, in particular, has useful applications in a variety of studies. The interpretation of “same”-“different” reaction time data, however, is complicated by the existence of two seemingly anomalous effects, the fast-“same” effect and the criterion effect. This article examines explanations of these effects and brings evidence from the literature to bear on them. Current models of “same”-“different” judgments, two of which are widely accepted, can be rejected on the evidence of data reviewed here. Among the disconfirming data are several results that offer suggestions both for improving current models and for questioning some of their fundamental presuppositions.

Effect of identity in the multiletter matching task

Article

Full-text available

Aug 1976

David A. Taylor

80 undergraduates took part in a study concerned with the nature of the character-comparison process involved in same-different judgments of arrays of letters. One condition replicated the standard matching task, with Ss pressing 1 key to indicate that the 2 arrays of letters were identical in all positions, and another key to indicate difference in 1 or more positions. A 2nd condition reversed the standard decision rule and required Ss to distinguish between arrays that differed in every position from those that matched in 1 or more positions. Reaction times were shorter to identical arrays in both conditions, indicating that identity effects observed in this and previous studies must be attributed to identity per se and not to decision or response factors. Mathematical analysis of the data revealed that, although a serial self-terminating model could be modified to give a good description of processing in this task, a model based on parallel processing was able to provide a better account both quantitatively and qualitatively. The analysis also indicated that Ss reduced their average reaction times by terminating the comparison process prematurely on some trials, and the guessing strategy had important implications for the shapes of the reaction time functions observed in this task. (15 ref)

Sequential determinants of information processing in serial and discrete choice reaction time

Article

Full-text available

Mar 1969

Sylvan Kornblum

It is shown that the measure of average stimulus information (H) is confounded with the probability of nonrepetition of the stimuli in most of the experimental conditions whose results have been taken as evidence in support of the linear relationship between choice reaction time (RT) and H. The results of a serial and a discrete experiment, so designed as to unconfound these two variables, lead to a rejection of the information hypothesis. The RT for repetitions is found to be faster than for nonrepetitions, and both are decreasing linear functions of their respective conditional probabilities. Some of the discussion focuses on the manner in which the slope and the intercepts of these linear functions are affected by changes in the number of alternatives, stimulus-response compatibility , and response-to-stimulus interval. It is also shown that the present approach not only accounts for data which had previously been described by the information hypothesis, but for results which departed from the hypothesis as well. Errors are discussed in a manner which supplements the main argument. Finally, it is shown that the molar results of RT experiments can be systematically accounted for in terms of the characteristics of easily distinguishable differentially sensitive partitions in these data.

The Fast–Same Effect of an Exclusive-OR Task

Article

Full-text available

Jun 2020

Participants are faster to decide that two stimuli are identical than to decide that they are different. Opposing theories suggested that this fast-same effect is either due (a) to a response bias toward similarity or (b) to facilitation caused by the repetition of the stimuli attributes. Although both theories predict the fast-same effect in a conventional same-different task, they make distinct predictions for tasks in which response bias is removed. In such tasks, the bias theory predicts that the fast-same would disappear whereas the facilitation theory predicts that the fast-same would remain. We tested those hypotheses using a same-different task in which participants had to indicate if all the attributes of the stimuli were matching or all were mismatching by pressing one response key, or if some attributes were matching and some were mismatching, by pressing another response key. We call this an exclusive-OR same-different task. Results show that participants were much faster in the "all-matching" condition compared with the "all-mismatching" condition, therefore supporting the facilitation theory. A fit of the linear ballistic accumulator model to the observed data provide additional supports that the fast-same effect is not caused by bias, but by a faster accumulation rate of evidence in the "all-matching" condition. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

The EZ Diffusion Model: An overview with derivation, software, and an application to the Same-Different task

Article

Full-text available

Apr 2020

The diffusion model is useful for analyzing data from decision making experiments as it gives information about a dataset that regular statistical tests cannot, including: the rate of processing, the encoding and motor response times, and decision thresholds. The EZ diffusion model is a restricted version of the diffusion model with some parameter variability set to zero, allowing for quicker analyses. Here we describe the EZ diffusion model --including how it was derived mathematically-- the measurement units of the parameters, and how it can be generalized to starting points other than the mid-point. We also show how its parameters can be estimated using computer software (the model is available with many software programs such as R and Excel, to which we add SPSS and a Mathematica code). Finally, an EZ analysis was run on one dataset obtained from a 'Same'-'Different' experiment.

Into the Mind's Eye: Exploring the Fast-Same Effect in the Same-Different Task

Article

Full-text available

Dec 2019

The fast-same effect is the observation that “same” responses are much faster than “different” responses in the same-different task. Moreover, identical stimuli are responded to faster than stimuli that are the same in name only (e.g., B and b). We examine Bamber's (1969) identity reporter model (a two-stage model predicting load effects), Proctor's (1981) facilitation framework, and Krueger and Shapiro's (1981) priming framework, proposed to account for these effects. Facilitation and priming are strong for identical and repeated stimuli, for phonological associates, and, in principle, for any form of association. We thus manipulated two types of associations: nominal (changing the letter case, preserving phonology) and learned (matching arbitrary symbols to letters) associations and used extended training to see variations in load effects. We found that overall performance benefits from phonological information and from training, although training did not change load effects. Additionally, the results from a transfer phase show capacity limitation for learned associates, which severely constrains facilitation. This finding is inconsistent with a priming framework. The results are discussed using an expanded version of Bamber's identity reporter model, which is also compatible with Proctor's facilitation framework.

Sequential Sampling Models of Same-Different Data and How They Explain the Fast-Same Effect

Article

Full-text available

Nov 2019

There have been many empirical endeavors to explain the fact that participants correctly identify same stimuli faster than different stimuli. Despite a large body of literature, a conclusive and unanimous consensus to explain this effect is lacking. The present study offers a new perspective on the same-different task using best-fitting parametric values of 2 sequential sampling models (linear ballistic accumulator [LBA] and drift-diffusion model [DDM]) to evaluate the contribution of 3 theoretical concepts (starting point, accumulation rate, and base time) to the fast-same effect. Eighteen participants were recruited to perform 3 tasks in which they compared letter strings (a) in a normal setting, the control task; (b) with lower contrast, the contrast task; and (c) based on nominal identity, the case task. Both the LBA and the DDM suggests that participants have faster base time in the same condition compared with the different condition. They also suggest that the number of letters and the number of differences influence the accumulation rate. These results suggest that the cognitive processes do not accumulate information as efficiently when the workload is increased. Finally, the limited predictive power of the best-fitting models suggests that better theoretical grounds are needed to identify fundamental cognitive concepts underlying same-different judgments. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

On the Time it Takes to Tell Things Apart

Chapter

May 2022

Raymond S. Nickerson

How many decimals? Rounding descriptive and inferential statistics based on measurement precision

Article

Aug 2020

Denis Cousineau

Reporting descriptive statistics requires rounding the results. Experienced researchers typically round the numbers to one or two decimals, following the APA manual. However, this general recommendation ignores the sample size and the instrument’s precision. Herein, expressions are derived that indicate how many decimals are reliable and so at what point the results should be rounded. The derivations are based on the measurement precision, that is, the precision of the raw data and uses propagation of error techniques. Two scenarios are considered, one in which systematic measurement error is possible and one in which only non-systematic measurement error is assumed. An example is provided showing how to round means, standard deviations and t values, among others, and an R library is available to automatize calculations.

The Power of Replicated Measures to Increase Statistical Power

Article

Jun 2019

When running statistical tests, researchers can commit a Type II error, that is, fail to reject the null hypothesis when it is false. To diminish the probability of committing a Type II error (β), statistical power must be augmented. Typically, this is done by increasing sample size, as more participants provide more power. When the estimated effect size is small, however, the sample size required to achieve sufficient statistical power can be prohibitive. To alleviate this lack of power, a common practice is to measure participants multiple times under the same condition. Here, we show how to estimate statistical power by taking into account the benefit of such replicated measures. To that end, two additional parameters are required: the correlation between the multiple measures within a given condition and the number of times the measure is replicated. An analysis of a sample of 15 studies (total of 298 participants and 38,404 measurements) suggests that in simple cognitive tasks, the correlation between multiple measures is approximately .14. Although multiple measurements increase statistical power, this effect is not linear, but reaches a plateau past 20 to 50 replications (depending on the correlation). Hence, multiple measurements do not replace the added population representativeness provided by additional participants.

Is the Fast-Same Phenomenon That Fast? An Investigation of Identity Priming in the Same-Different Task

Abstract and Figures

Recommended publications

The Fast–Same Effect of an Exclusive-OR Task

Analyses of Response Time Data in the Same–Different Task

The cognitive architecture of processes responsible to assess similarity and clarity in a comparison...

Into the Mind's Eye: Exploring the Fast-Same Effect in the Same-Different Task