Behavior Analysis in Practice (2021) 14:679–703

https://doi.org/10.1007/s40617-020-00527-7

PRECISION TEACHING: DISCOVERIES AND APPLICATIONS

An Evaluation of Multiple SAFMEDS Procedures

Shawn P. Quigley 1,2 & Stephanie Peterson 1 & Jessica E. Frieder 1 & Kimberly M. Peck 1 & Anthony Kennedy-Walker 1 &
Meaghan Chirinos 2

Accepted: 18 October 2020 / Published online: 2 June 2021

# Association for Behavior Analysis International 2021

Abstract
Lindsley developed the “say all fast minute every day shuffled” (SAFMEDS) procedure in the late 1970s to enhance the typical
use of flash cards (Graf & Auman, 2005). The acronym specifically guides the learner’s behavior when using flash cards. A
review of SAFMEDS research indicates its successful use with children, college students, and older adults with and without
disabilities. The literature also indicates that SAFMEDS procedures are not well documented and have multiple variations,
limiting practitioners’ ability to know what procedures to use and when. The purpose of this study was to evaluate the effects
of a basic SAFMEDS procedure and four supplementary SAFMEDS procedures on the rates of correct and incorrect responding
to unfamiliar Russian words and Chinese characters in college students. The results of the study suggest that the basic SAFMEDS
procedure produced some learning (i.e., increases in correct responding and decreases in incorrect responding), but all of the
supplementary procedures led to greater increases in the number of correct responses per 1-min timing. Further research
evaluating differences in performance across the supplementary procedures is warranted.

Keywords SAFMEDS . Flash cards . Precision teaching . Self-study

Precision teaching (PT) is a supplemental strategy emphasizing structured and do not necessarily emphasize important areas
the frequent measurement of student performance, which in turn of instruction, such as making an observable response (i.e.,
allows teachers to make instructional decisions (Merbitz, Vieitez, saying the answer). It is difficult to determine the “gold stan-
Merbitz, & Pennypacker, 2004; West, Young, & Spooner, dard” SAFMEDS procedure because so many procedural var-
1990). Hallmarks of PT are the frequent measurement of learn- iations exist. However, Eshleman (2000a) and Graf and
ing, standard graphing of data, analysis of data to guide instruc- Auman (2005) described a basic procedure for SAFMEDS
tional decisions, and instructional changes that maximize learn- that is often referred to in the literature. Their procedure con-
ing (West et al., 1990). One strategy linked to PT is “say all fast sists of the following steps, in which the learner
minute every day shuffled” (SAFMEDS).
SAFMEDS is an instructional and assessment strategy sim- 1. Holds the deck,
ilar to flash cards. Lindsley coined the term in the late 1970s to 2. Shuffles the cards,
specify procedural differences between SAFMEDS and tradi- 3. Starts the timer (typically 1 min),
tional flash card procedures (Eshleman, 2000a; Graf & 4. “Sees” the front of the card and “says” the term on the
Auman, 2005; Lindsley, 1996). Flash card procedures are less back of the card out loud (a see/say learning channel,
although other learning channels can be used; see Graf
This project was completed in partial fulfillment of Shawn P. Quigley’s
& Auman, 2005, for further discussion),
PhD requirements at Western Michigan University. 5. Turns the card over and checks the answer for immediate
feedback,
* Shawn P. Quigley 6. Sorts the cards for correct and incorrect responses into
[email protected] piles,
7. Counts the number of responses per pile after the time
1
Psychology Department, Western Michigan University, MI expires, and
Kalamazoo, USA 8. Charts their own performance for review and instructional
2
Present address: Melmark, Pennsylvania, 2600 Wayland Rd., changes (see Calkin, 2005, for further information regard-
Berwyn, PA 19312, USA ing a standard celeration chart [SCC]).
680 Behav Analysis Practice (2021) 14:679–703

The simplest SAFMEDS procedure (based strictly on the the basic procedure described previously. Further, 11 of the
guidelines suggested within the acronym) involves conducting publications did not contain a sufficient description of the
a single timing per day using the aforementioned procedures. SAFMEDS procedures to ascertain what procedure was imple-
However, Eshleman and Graf and Auman provided in their mented. Of the 16 studies with a description of the SAFMEDS
how-to guides other examples of how SAFMEDS instruction procedures, the procedures varied along the parameters of the
and assessment might be implemented. They suggested that number timings conducted per day, whether the learner had the
multiple timings per day could be used instead of just one, opportunity to review the cards prior to the timing, the length of
the number of cards within the deck could be increased and the timings (e.g., 30 s, 1 min), the total number of cards in a
decreased based on learner performance, and the length of the deck, and the error correction procedures used. Quigley et al.
timing could be increased or decreased. also noted that researchers also occasionally implemented
Since its inception, SAFMEDS has been evaluated to de- multiple SAFMEDS procedures within the same research
termine its effects on learning (e.g., Bolich & Sweeney, 1996; project. For example, Beverly, Hughes, and Hastings (2009)
Byrnes, Macfarlane, Young, & West, 1990; Eaton & Fox, added additional timings, altered the total number of cards
1983; Eshleman, 1985) and the maintenance of learning reviewed, and introduced error correction procedures based
(Kim, Carr, & Templeton, 2001; Olander, Collins, on student performance, all within the same study. A clear
McArthur, Watts, & McDade, 1986). In addition, compari- understanding of intervention procedures is critical for any sort
sons between different SAFMEDS procedures (e.g., teacher of discussion of validity. Yet, all of the procedural variations
vs. student deck development; McDade & Olander, 1990) and found in the literature make it difficult to determine what the
modalities (e.g., computer-based vs. paper deck; McDade, effects of various SAFMEDS procedures are and when to uti-
Austin, & Olander, 1985) have been made, as well as between lize each variation to maximize learning.
factors affecting performance outcomes (e.g., error rate; According to Kazdin (2003), internal validity is the ability to
Bower & Orgel, 1981; stimulus control; Meindl, Ivy, Miller, rule out explanations of results other than the designated inter-
Neef, & Williamson, 2013). A review of this literature indi- vention. Additionally, Kazdin (2003) stated that external valid-
cates that SAFMEDS increases the number of correct answers ity is the extent to which the results are applicable to other
across varied content (e.g., math facts, reading fluency, posi- populations, settings, and circumstances beyond the research.
tive self-statements), various populations (i.e., elementary ed- A clear description of the intervention is necessary to evaluate
ucation, secondary education, university, and geriatric), and potential confounding variables for internal validity.
various settings (e.g., education, home, and community; Furthermore, a clear description is needed when evaluating
Quigley, Peterson, Frieder, & Peck, 2017). how the intervention might be applicable to other populations,
Anecdotally, we have used SAFMEDS in our own instruc- settings, and circumstances beyond the research. Internal and
tion with college students, and we are aware of others who do external validity is increased when studies are replicated across
so as well. It is not uncommon for instructors to “tinker” with time using the same intervention procedures. Clear descriptions
the SAFMEDS procedure to meet the needs of students. For of the intervention impact our ability to establish internal and
example, our own students have expressed that it is over- external validity, which in turn impacts effective practice.
whelming to be presented with a large deck of cards at the Currently, the literature base on SAFMEDS provides an
beginning of a semester that must be mastered by the end of unclear description of the SAFMEDS procedure, making it
the semester. As a result, we have divided the cards into sets so difficult for a practitioner to determine what constitutes
that the students could work on smaller subsets of the cards. evidence-based practice when implementing them. Research
Sometimes, these subsets are combined as the semester pro- is needed to clearly identify how different variations of the
gresses so that students have an increasing number of cards in SAFMEDS procedure impact learning. Thus, the purposes
their decks. In addition to making the task less daunting at the of this study were to clearly define multiple SAFMEDS pro-
onset of the semester, this procedure also has the benefit of not cedures and evaluate the effects of various SAFMEDS proce-
introducing cards for concepts that have not yet been taught in dures on the correct and incorrect responses to vocabulary
the semester. This “tinkering” with the procedure has pro- terms of a novel language.
duced a literature base that is not entirely clear as to what
the “standard” SAFMEDS strategy is. Furthermore, it is un-
clear what modifications to the procedure can be made, when Method
those modifications should be made, and whether those mod-
ifications reliably produce better learning. Participants
To illustrate, Quigley et al. (2017) reviewed 27 SAFMEDS
publications and noted the SAFMEDS procedures implement- Nineteen participants enrolled at a midwestern university were
ed by researchers. Of the 27 publications, none of the re- recruited for participation in the study through flyers and in-
searchers utilized a SAFMEDS procedure that aligned with class recruitment scripts. Participants had varied educational
Behav Analysis Practice (2021) 14:679–703 681

backgrounds (e.g., GPA, enrollment status), varied exposure one side (e.g., a picture of a car) and a corresponding common
levels to SAFMEDS, and varied histories of exposure to term on the other side (e.g., “car”). In an attempt to train
Chinese and Russian languages. Criteria for inclusion in the participants in the procedures for when they made an incorrect
study included the participants having basic reading skills, response, some cards had a common picture on one side (e.g.,
vocal language, no physical disabilities that limited hand/ a picture of an 18-wheeler truck) with an obscure correspond-
finger dexterity, no visual impairments, and no history of ing term on the other side (e.g., “big rig”). Last, in an attempt
learning related to the target languages (i.e., the participants to train participants in the procedures for when they wanted to
had no formal training in the Chinese or Russian language, “pass” on a response, some cards had an unfamiliar and ob-
neither the Chinese nor Russian language was spoken at scure picture on one side (e.g., a picture of a dragon fruit) with
home, and the participant had not resided in a location where the corresponding term on the other side (e.g., “dragon fruit”).
the Russian or Chinese language was primarily spoken). The The Russian and Chinese decks were created in a similar
inclusionary criteria were assessed via a questionnaire. All 19 manner and differed from the training deck only in content
volunteers met these inclusionary criteria and enrolled in the and total number of cards. One deck had Mandarin Chinese
study. However, only 14 of the volunteers completed the characters on one side with the English translation on the
study. Attrition was largely due to conflicts with the time opposite side. The other deck had Russian words on one side
commitment of daily research sessions. See Table 1 for addi- with the English translation on the opposite side. Both decks
tional participant demographics. contained 60 cards with the same English words across decks
(e.g., 月 = moon and Луна = moon; 山 = mountain and ropa =
mountain). Correspondence between languages was deter-
Settings and Materials mined via Google Translate (http://translate.google.com/).
All information on the fronts and backs of the cards
Sessions were conducted in a small room at the university. appeared in the center of the cards. To avoid stimulus
The room measured approximately 3 m × 2 m and contained control issues such as smudge marks or bent corners
tables and chairs. The participant and one or two researchers evoking responses instead of the printed words (see
were present for every session; a second researcher was occa- Eshleman, 2000a, 2000b, and Meindl et al., 2013, for further
sionally present for live scoring of interobserver agreement discussion), multiple identical decks were created and rotated
(IOA) data. Materials consisted of two timers (i.e., stopwatch throughout sessions. Cards with any visible marks, smudges,
and digital countdown time), three different decks of tears, or other defects were replaced with new cards.
SAFMEDS cards (i.e., training deck, Chinese deck, and
Russian deck), a demographics questionnaire for assessing
inclusionary criteria, and video-recording equipment. All ses- Dependent Variables and Data Analysis
sions were video recorded for later data review purposes.
The SAFMEDS cards were approximately 7.6 cm × The dependent variables were the number of correct responses
12.7 cm and were created by printing the relevant information and incorrect responses participants made to the cards in each
on both sides of card stock and then cutting the cards out. of the decks. A correct response occurred when the participant
There were three decks of cards: a training deck, a Russian stated out loud the term associated with the character/word on
deck, and a Chinese deck. The training deck consisted of 20 the other side of the card. For example, if the participant saw
cards, each with a picture on one side and a term on the other. the Chinese character for “boy,” they vocally stated “boy” for
The picture/term combinations were created to train partici- a correct response. An incorrect response occurred when the
pants on how to complete the SAFMEDS procedure. In an participant stated a term other than what was printed on the
attempt to train participants in the procedures for when they card. For example, if the participant saw the Chinese character
made a correct response, some cards had a common picture on for “boy” and they said “girl,” this was counted as incorrect. If

Table 1 Participant Demographics

Education information Language history SAFMEDS exposure

BA/ Master’s PhD GPA < 3.0 GPA > 3.0 Second-language training Years of training < 2 Years of training > 2 Total number
BS

8 5 1 2 12 12 3 9 9

Note. Each number reflects the total number of participants meeting the demographic criteria. The degree reflects what degree level the participant was
enrolled in at the time of participation.
682 Behav Analysis Practice (2021) 14:679–703

the participant provided more than one response for a single cards prior to the timing). All participants experienced the basic
card, the final response was scored according to the aforemen- SAFMEDS procedure for both the Chinese and Russian decks.
tioned directions. For example, if the participant saw the Each participant then experienced one of the four modified
Chinese character for “boy” and said “Girl . . . no, boy,” it SAFMEDS procedures for one of the decks while continuing
was counted as correct so long as the final response occurred the basic SAFMEDS procedure for the other deck. For exam-
prior to seeing the back side of the card. Participants could ple, Participant 1 first experienced daily 1-min timings using the
state “pass” if they were unsure of the correct answer. If the basic procedure for both the Chinese and Russian decks for 2
participant passed, they still flipped the card over to see the weeks. At the start of the 3rd week, Participant 1 continued
correct response. We counted passes as incorrect responses. using the basic procedure for the Russian deck but experienced
There were no time constraints within the 1-min timing to the whole-deck procedure for the Chinese deck. The selection
produce a response for each individual card. The primary ob- of which additional SAFMEDS procedure each participant ex-
server watched video recordings to record responses. perienced was randomly determined. Table 2 provides a sum-
Metrics for level, celeration, and bounce were calculated mary of each independent variable; more detail is provided in
using the statistical software PrecisionX (CentralReach, the descriptions of each independent variable. In the following
2020). These values were analyzed in combination with visual descriptions, the testing timing refers to the 1-min timing con-
analysis. ducted as part of the basic procedure. For any supplemental
procedures that included other timings (e.g., sprints), the testing
Interobserver Agreement timing refers to the final 1-min timing that followed the practice
timings. We only report the results of the testing timing for each
IOA data were recorded by having a second observer inde- participant.
pendently watch the videos and record participant responses
to the cards. IOA data were calculated using a point-by-point Basic SAFMEDS Procedure The basic SAFMEDS procedure
agreement method (Kazdin, 1982) across a minimum of 25% was based on the procedures described by Eshleman (2000a)
of sessions, distributed equally across all phases for each par- and Graf and Auman (2005). Participants were told,
ticipant. Overall agreement averaged 99% (range 89%–100%)
across all participants and phases of the study.
You will have 1 minute to say as many terms as you can.
Independent Variables If you don’t know the term, you can pass on the card. If
you pass, you should still flip the card over to see the
Five independent variables were implemented and evaluated. correct response.
The first independent variable was a package of variables that
composed the basic SAFMEDS procedure. The other four in- Participants then held the deck, shuffled the deck, started a
dependent variables consisted of this basic procedure along countdown timer set for 1 min, looked at the stimulus on the
with one variation of the procedure that had been documented front side of the card (i.e., see), provided a vocal response (i.e.,
in Quigley et al. (2017). In these cases, all aspects of the pro- say the answer or say “pass”), flipped the card over to expose
cedure were the same as the basic procedure except for the the correct answer and to determine whether the stated answer
variable that was modified. These four modifications consisted was correct or incorrect, and placed the card onto a corre-
of a whole-deck presentation of cards, an incremental whole- sponding pile (i.e., correct, incorrect, or pass). The participants
deck presentation of cards, three 1-min sprints prior to timings, repeated this procedure with as many cards as possible during
and incremental sprints (three 1-min timings with a subset of the 1-min timing.

Table 2 Description of IV Components

IV Material review outside Time for outside Practice timings? Number of practice timings All cards in Test time
of test time? review before the test timing the deck?

Basic No No Yes 60 s
Whole deck Yes ≤10 m No Yes 60 s
Incremental whole deck Yes ≤10 m No No 60 s
Sprints No Yes 3 Yes 60 s
Incremental sprints No Yes 3 No 60 s

Note. IV = independent variable. If a cell is blank, the components were not present for that independent variable
Behav Analysis Practice (2021) 14:679–703 683

Whole-Deck Procedure This procedure was based on the pro- After the participant had experienced each subdeck, the 15
cedures described by Eshleman (2000a). Participants were cards from each deck were combined together (i.e., 60 cards,
provided up to 10 min to study the cards in any manner. If each containing a different word pair). Participants were told,
the participant studied the cards for the entire 10 min, the
researcher stopped the participant and conducted the 1-min
test timing. The participant could stop the 10-min study period You will have 1 minute to say as many terms as you can.
at any point and proceed to the 1-min test timing. Participants If you don’t know the term, you can pass on the card. If
were told, “Review each card as quickly as possible. If your you pass, you should still flip the card over to see the
review continues for 10 minutes, you will be stopped.” correct response.

Incremental Whole-Deck Procedure This procedure was based

on the procedures described in Kim et al. (2001). The incre- Fidelity of the Independent Variable
mental whole-deck procedure was identical to the whole-deck
procedure (i.e., study time followed by test timing), except A fidelity checklist was used to determine if key procedural
that participants were only given a portion of the cards from aspects of each SAFMEDS procedure (e.g., shuffling, number
the deck. Rather than being given all 60 cards in the deck, of timings, duration of timing, sorting of cards into correct
participants were given a subdeck of 15 cards at a time. piles) were followed across sessions and participants.
However, 4 duplicates of each card were provided so that a Fidelity was assessed for at least 25% of the sessions across
total of 60 cards were still presented. The cards for the test all phases. IOA data were also taken on fidelity measures by
timing were the same ones used during the practice period. having a second observer simultaneously but independently
Subdecks containing novel cards (i.e., 15 new cards with 4 complete the same fidelity checklist. All fidelity checks were
duplicates) were introduced at different points. After the par- conducted by watching video recordings of sessions. Fidelity
ticipant had experienced each subdeck, the 15 cards from each IOA was calculated using a point-by-point agreement method
deck were combined together (i.e., 60 cards, each containing a (Kazdin, 1982). Fidelity averaged 98% (range 80%–100%)
different word pair). Participants were told, “Review each card across all participants and phases. Fidelity IOA averaged
as quickly as possible. If your review continues for 10 mi- 99% (range 80%–100%) across all participants and phases.
nutes, you will be stopped.”
Experimental Design and Procedures
Sprints Procedure This procedure was basedupon procedures
described by Eshleman (2000a) and Stockwell and Eshleman The research questions were evaluated using a simultaneous-
(2010). Participants practiced the word pairs by conducting treatment design embedded within in a concurrent multiple-
three 1-min timings following the basic procedure (described baseline across-participants design (Kazdin, 1982). That is,
in the introduction), using the full deck. Each timing was each participant experienced two variations of SAFMEDS
completed back-to-back with less 1-min between timings. simultaneously, one with each deck (Chinese and Russian).
Following the three practice timings, a fourth timing, the test However, the implementation of the variation strategy was
timing, was conducted. Participants were told, staggered across participants. The simultaneous-treatment de-
sign allowed for within-subject comparisons of the basic
SAFMEDS procedure and one of the supplementary
You will have 1 minute to say as many terms as you can. SAFMEDS procedures. The multiple-baseline design provid-
If you don’t know the term, you can pass on the card. If ed the opportunity for replication of effects across participants
you pass, you should still flip the card over to see the and SAFMEDS procedures. The multiple baselines were
correct response. discontinuous.

Basic SAFMEDS Procedure Training Participants were trained

Incremental Sprints Procedure This procedure was an exten- on the basic SAFMEDS procedure with the training deck.
sion of the procedures in Kim et al. (2001). The procedure was Participants were told how to hold the deck, shuffle the deck,
identical to the sprints procedure, except that the participant start the countdown timer, look at the stimulus on the front
was only given a portion of the cards from the deck. As with side of the card (i.e., see), provide a vocal response (i.e., say),
the incremental whole-deck procedure, participants were only flip the card over to determine whether the answer was correct
given 15 cards at a time with 4 duplicates of each of the 15 or incorrect, place the card into a pile (i.e., correct, incorrect,
cards for a total of 60 cards. There were four different subdecks or pass), and repeat the procedure with as many cards as pos-
(i.e., each with 15 cards with 4 duplicates) and the complete sible during the 1-min timing. In addition to vocal instruction,
deck (i.e., 60 cards, each containing a different word pair). participants watched a video demonstration of the basic
684 Behav Analysis Practice (2021) 14:679–703

SAFMEDS procedure, and the first researcher modeled the demonstrate real-time changes in day-to-day performance
procedure. Next, participants were asked to implement the for each participant.
SAFMEDS procedure. Feedback was provided for correct
and incorrect implementation. This was repeated Whole Deck
until participants made at least 20 responses per minute with
80% or higher fidelity. All participants met these criteria in Level Across the three participants in the whole-deck supple-
four or fewer practice timings. mental SAFMEDS procedure, participants demonstrated an
average of 12.62 (range 7.35–15.62) correct responses per
Basic SAFMEDS Procedure for Chinese and Russian Decks day throughout the duration of this procedure (see Table 3).
Following successful completion of training, participants Incorrect responses in the whole-deck procedure occurred at
met daily (if possible) with a researcher to complete test tim- an average rate of 9.61 (range 6.59–11.71) per day across all
ings using the basic SAFMEDS procedure described previ- sessions. In comparison, in the basic SAFMEDS procedure,
ously for each of the Chinese and Russian decks. Participants these same participants yielded an average rate of 1.88 (range
only had access to the cards during the 1-min timing. This 1.27–2.53) correct responses per day. The basic SAFMEDS
served as a baseline for both decks for each participant. procedure showed an average of 20.51 (range 17.8–23) incor-
rect responses per day. In general, correct response rates were
Introduction of Additional SAFMEDS Procedure In a staggered low in the basic SAFMEDS conditions, especially when com-
fashion across participants, participants then experienced one pared to incorrect responses.
of the four variations of the SAFMEDS procedure.
Participants were randomly assigned to a procedure based Celeration Celerations for the number of correct responses in
on their order of enrollment in the study. The SAFMEDS the whole-deck procedure ranged from ×1.27 (27% accelera-
variation was introduced for only one of the decks, whereas tion in daily corrects over time) to ×1.63 (63% acceleration in
the participant continued to use the basic SAFMEDS proce- daily corrects over time), with an average of ×1.5. Celerations
dure for the other deck. First, participants were trained on the for the number of incorrect responses in the whole-deck pro-
new procedure using the same strategies as were used for the cedure ranged from ÷1.1 (10% decay in incorrects) to ÷1.4
training deck. The deck for which the participant used the (40% decay in incorrects), with an average of ÷1.21. In com-
SAFMEDS variation (i.e., Russian or Chinese) was parison, celerations for the number of correct responses in the
counterbalanced across participants. basic SAFMEDS procedure ranged from ×1.0 (no daily in-
crease in corrects over time) to ×1.33 (33% acceleration in
Follow-Up Phase The purpose of this phase was to briefly daily corrects over time), with an average of ×1.22.
assess the short-term maintenance of performance following Celerations for the number of incorrect responses in the basic
at least 10 days without access to the cards. Following at least SAFMEDS procedure ranged from ÷1.0 (no decay in incor-
a 10-day break, participants completed a single 1-min timing rects over time) to ×1.05 (5% acceleration in daily incorrects
for each deck following the basic SAFMEDS procedure. over time), with an average of ×1.03.

Bounce The bounce of correct responses in the whole-deck

Results procedure ranged from ×2.4 (140% acceleration in the
variability of corrects over time) to ×3.6 (260% accelera-
Results for participants are provided in Figs 1, 2, 3, 4, 5, 6, 7, tion in the variability of corrects over time), with an av-
8, 9, 10, 11, 12, 13 and 14, with summary data presented in erage of ×2.9. The bounce of incorrect responses in the
Tables 3, 4, 5 and 6. Figures are grouped by participant and whole-deck procedure ranged from ×1.5 (50% accelera-
SAFMEDS procedure (i.e., the first figure is the basic tion in the variability of incorrects over time) to ×2.3
SAFMEDS procedure, and the second figure is the supple- (130% acceleration in the variability of incorrects over
mental procedure for the participant); the tables summarize time), with an average of ×1.9. In comparison, the bounce
the comparative metrics of each supplemental SAFMEDS of correct responses in the basic SAFMEDS procedure
procedure with the basic SAFMEDS procedure across cor- ranged from ×3.6 (260% acceleration in the variability
rects and incorrects for each participant. Figures represent of corrects over time) to ×4.3 (330% acceleration in the
the number of correct and incorrect (i.e., including pass re- variability of corrects over time), with an average of
sponses) responses per 1-min timing across days. The duration ×4.03. The bounce of incorrect responses in the basic
of baseline was staggered across participants to create a dis- SAFMEDS procedure ranged from ×1.3 (30% accelera-
continuous multiple-baseline across-participants design; how- tion in the variability of incorrects over time) to ×1.6
ever, this may not be evident due to the presentation of charts (60% acceleration in the variability of incorrects over
by participant. Results are depicted using SCCs to time), with an average of ×1.4.
Behav Analysis Practice (2021) 14:679–703 685

Karly

Fig. 1 Results for Karly. Note. The top chart shows the raw data for the basic procedure using the Chinese deck. The bottom chart shows the raw data for
the whole-deck procedure using the Russian deck

Incremental Whole Deck (Bailey is not included in the all-deck calculations due to
limited data). Incorrect responses across subdecks occurred
Level Across the four participants in the incremental whole- at an average rate of 1.94 (range 0.50–7.63) per day across
deck supplemental SAFMEDS procedure, participants dem- all sessions. Incorrect responses using the entire deck occurred
onstrated an average of 28.91 (range 14.59–37.51) correct at an average rate of 5.76 (range 2.57–14) per day across all
responses per day across subdecks (see Table 4). sessions. In comparison, in the basic SAFMEDS procedure,
Participants demonstrated an average of 20.20 (range 12– these same participants yielded an average rate of 1.89 (range
27.31) correct responses per day using the entire deck 0.60–3.02) correct responses per day. The basic SAFMEDS
686 Behav Analysis Practice (2021) 14:679–703

Terri

Fig. 2 Results for Terri. Note. The top chart shows the raw data for the basic procedure using the Russian deck. The bottom chart shows the raw data for
the whole-deck procedure using the Chinese deck

procedure yielded an average of 23.16 (range 10.76–33.84) time) to ×10.2 (920% acceleration in daily corrects over time),
incorrect responses per day. In general, correct response rates with an average of ×2.72. Celerations for the number of cor-
were low in the basic SAFMEDS conditions, especially when rect responses using the entire deck ranged from ×1.07 (7%
compared to incorrect responses. acceleration in daily corrects over time) to ×1.52 (52% accel-
eration in daily corrects over time), with an average of ×1.29.
Celeration Celerations for the number of correct responses in Celerations for the number of incorrect responses, across all
the incremental whole-deck procedure, across all subdecks, subdecks, ranged from ×1.0 (no change in incorrects) to ÷128
ranged from ×1.06 (6% acceleration in daily corrects over (1,280% decay in incorrects), with an average of ÷16.56.
Behav Analysis Practice (2021) 14:679–703 687

Louise

Fig. 3 Results for Louise. Note. The top chart shows the raw data for the basic procedure using the Chinese deck. The bottom chart shows the raw data
for the whole-deck procedure using the Russian deck

Celerations for the number of incorrect responses using the incorrect responses in the basic SAFMEDS procedure ranged
entire deck ranged from ÷1.2 (20% decay in daily corrects from ×1.05 (5% acceleration of incorrects over time) to ÷1.03
over time) to ÷6.16 (516% decay in daily incorrects over (3% decay in daily incorrects over time), with an average of
time), with an average of ×2.99. In comparison, celerations ×1.03.
for the number of correct responses in the basic SAFMEDS
procedure ranged from ×1.09 (9% daily increase in corrects Bounce The bounce of correct responses in the incremental
over time) to ×1.18 (18% acceleration in daily corrects over whole-deck procedure, across all subdecks, ranged from ×1.0
time), with an average of ×1.13. Celerations for the number of (no daily change in the variability of corrects over time) to
688 Behav Analysis Practice (2021) 14:679–703

Jackson

Fig. 4 Results for Jackson. Note. The top chart shows the raw data for the basic procedure using the Russian deck. The bottom chart shows the raw data
for the incremental whole-deck procedure using the Chinese deck

×2.5 (150% acceleration in the variability of corrects over of corrects over time) to ×8.1 (710% acceleration in the vari-
time), with an average of ×1.3. The bounce of correct re- ability of incorrects over time), with an average of ×2.5. The
sponses using the entire deck ranged from ×1.3 (30% accel- bounce of incorrect responses using the entire deck ranged
eration in the variability of daily corrects over time) to ×1.8 from ×2.8 (180% acceleration in the variability of daily incor-
(80% acceleration in the variability of daily corrects over rects over time) to ×3.0 (200% acceleration in the variability
time), with an average of ×1.63. The bounce of incorrect re- of daily incorrects over time), with an average of ×2.8. In
sponses in the incremental whole-deck procedure, across all comparison, the bounce of correct responses in the basic
subdecks, ranged from ×1.0 (no daily change in the variability SAFMEDS procedure ranged from ×3.6 (260% acceleration
Behav Analysis Practice (2021) 14:679–703 689

Bailey

Fig. 5 Results for Bailey. Note. The top chart shows the raw data for the basic procedure using the Chinese deck. The bottom chart shows the raw data for
the incremental whole-deck procedure using the Russian deck. Only one session of the final intervention phase (i.e., all cards) was completed

in the variability of corrects over time) to ×4.3 (330% Sprints

acceleration in the variability of corrects over time), with
an average of ×4.03. The bounce of incorrect responses in Level Across the four participants in the sprints supplemental
the basic SAFMEDS procedure ranged from ×1.3 (30% SAFMEDS procedure, participants demonstrated an average
acceleration in the variability of incorrects over time) to of 9.19 (range 1.72–17.77) correct responses per day through-
×1.6 (60% acceleration in the variability of incorrects out the duration of this procedure (see Table 5). Incorrect
over time), with an average of ×1.4. responses in the sprints procedure occurred at an average rate
690 Behav Analysis Practice (2021) 14:679–703

Joan

Fig. 6 Results for Joan. Note. The top chart shows the raw data for the basic procedure using the Russian deck. The bottom chart shows the raw data for
the incremental whole-deck procedure using the Chinese deck

of 27.05 (range 17.13–32.52) per day across all sessions. In Celeration Celerations for the number of correct responses in
comparison, in the basic SAFMEDS procedure, these same the sprint procedure ranged from ×1.27 (27% acceleration in
participants yielded an average rate of 1.82 (range 1.03–2.46) daily corrects over time) to ×1.48 (48% acceleration in daily
correct responses per day. The basic SAFMEDS procedure corrects over time), with an average of ×1.38. Celerations for
showed an average of 32.56 (range 28.14–39.45) incorrect the number of incorrect responses in the sprint procedure
responses per day. In general, correct response rates were ranged from ×1.05 (5% acceleration in incorrects) to ÷1.08
low in the basic SAFMEDS conditions, especially when com- (8% decay in incorrects), with an average of ×1.04. In com-
pared to incorrect responses. parison, celerations for the number of correct responses in the
Behav Analysis Practice (2021) 14:679–703 691

Pam

Fig. 7 Results for Pam. Note. The top chart shows the raw data for the basic procedure using the Chinese deck. The bottom chart shows the raw data for
the incremental whole-deck procedure using the Russian deck

basic SAFMEDS procedure ranged from ×1.12 (12% acceler- Bounce The bounce of correct responses in the sprint proce-
ation in daily corrects over time) to ×1.35 (35% acceleration in dure ranged from ×3.2 (220% acceleration in the variability of
daily corrects over time), with an average of ×1.23. corrects over time) to ×3.7 (270% acceleration in the variabil-
Celerations for the number of incorrect responses in the basic ity of corrects over time), with an average of ×3.47. The
SAFMEDS procedure ranged from ×1.03 (3% acceleration in bounce of incorrect responses in the sprint procedure ranged
incorrects over time) to ×1.08 (8% acceleration in daily incor- from ×1.2 (20% acceleration in the variability of incorrects
rects over time), with an average of ×1.04. over time) to ×1.9 (90% acceleration in the variability of
692 Behav Analysis Practice (2021) 14:679–703

Kit

Fig. 8 Results for Kit. Note. The top chart shows the raw data for the basic procedure using the Russian deck. The bottom chart shows the raw data for the
sprints procedure using the Chinese deck. Follow-up was not completed for Kit

incorrects over time), with an average of ×1.5. In comparison, Incremental Sprints

the bounce of correct responses in the basic SAFMEDS pro-
cedure ranged from ×3.1 (210% acceleration in the variability Level Across the three participants in the incremental sprints
of corrects over time) to ×4.7 (370% acceleration in the vari- supplemental SAFMEDS procedure, participants demonstrat-
ability of corrects over time), with an average of ×3.77. The ed an average of 14.11 (range 8.07–18.54) correct responses
bounce of incorrect responses in the basic SAFMEDS proce- per day across subdecks (see Table 6). Participants demon-
dure ranged from ×1.2 (20% acceleration in the variability of strated an average of 13.07 (range 10.14–14.69) correct re-
incorrects over time) to ×1.4 (40% acceleration in the variabil- sponses per day using the entire deck. Incorrect responses
ity of incorrects over time), with an average of ×1.3. across subdecks occurred at an average rate of 10.33 (range
Behav Analysis Practice (2021) 14:679–703 693

Diane

Fig. 9 Results for Diane. Note. The top chart shows the raw data for the basic procedure using the Chinese deck. The bottom chart shows the raw data for
the sprints procedure using the Russian deck

3.98–18.96) per day across subdecks. Incorrect responses day. In general, correct response rates were low in the basic
using the entire deck occurred at an average rate of 14.11 SAFMEDS conditions, especially when compared to incor-
(range 7.49–22.17) per day across all sessions. In comparison, rect responses.
in the basic SAFMEDS procedure, these same participants
yielded an average rate of 3.04 (range 1.38–4.74) correct re- Celeration Celerations for the number of correct responses in
sponses per day. The basic SAFMEDS procedure showed an the incremental sprints procedure, across all subdecks, ranged
average of 24.54 (range 20.45–29.42) incorrect responses per from ÷1.09 (9% decay in daily corrects over time) to ×89.8
694 Behav Analysis Practice (2021) 14:679–703

Myra

Fig. 10 Results for Myra. Note. The top chart shows the raw data for the basic procedure using the Chinese deck. The bottom chart shows the raw data
for the sprints procedure using the Russian deck

(8,880% acceleration in daily corrects over time), with an ÷106.58. Celerations for the number of incorrect responses
average of ×11.48. Celerations for the number of correct re- using the entire deck ranged from ÷1.09 (9% decay in daily
sponses using the entire deck ranged from ×1.05 (5% accel- corrects over time) to ×1.02 (2% acceleration in daily incor-
eration in daily corrects over time) to ×1.61 (61% acceleration rects over time), with an average of ×0.37. In comparison,
in daily corrects over time), with an average of ×1.31. celerations for the number of correct responses in the basic
Celerations for the number of incorrect responses, across all SAFMEDS procedure ranged from ×1.1 (10% daily increase
subdecks, ranged from ÷1.33 (33% decay in incorrects) to in corrects over time) to ×1.14 (40% acceleration in daily
÷907 (9,060% decay in incorrects), with an average of corrects over time), with an average of ×1.12. Celerations
Behav Analysis Practice (2021) 14:679–703 695

Kylie

Fig. 11 Results for Kylie. Note. The top chart shows the raw data for the basic procedure using the Russian deck. The bottom chart shows the raw data
for the sprints procedure using the Chinese deck

for the number of incorrect responses in the basic SAFMEDS to ×3.9 (290% acceleration in the variability of corrects
procedure ranged from ÷1.05 (5% decay of incorrects over over time), with an average of ×2.11. The bounce of cor-
time) to ×1.04 (4% acceleration in daily incorrects over time), rect responses using the entire deck ranged from ×1.6
with an average of ×1.0. (60% acceleration in the variability of daily corrects over
time) to ×2.4 (140% acceleration in the variability of daily
Bounce The bounce of correct responses in the incremental corrects over time), with an average of ×1.9. The bounce
sprints procedure, across all subdecks, ranged from ×1.4 of incorrect responses in the incremental whole-deck pro-
(40% acceleration in the variability of corrects over time) cedure, across all subdecks, ranged from ×1.3 (30%
696 Behav Analysis Practice (2021) 14:679–703

Nathan

Fig. 12 Results for Nathan. Note. The top chart shows the raw data for the basic procedure using the Russian deck. The bottom chart shows the raw data
for the incremental sprints procedure using the Chinese deck

acceleration in the variability of incorrects over time) to Discussion

×6.2 (520% acceleration in the variability of incorrects
over time), with an average of ×3.03. The bounce of in- SAFMEDS is an assessment and instructional strategy based
correct responses using the entire deck ranged from ×1.3 on PT principles (Eshleman, 2000a; Graf & Auman, 2005).
(30% acceleration in the variability of daily incorrects The name suggests a standard application for use, but a liter-
over time) to ×3.5 (250% acceleration in the variability ature review suggests vague and varied applications of the
of daily incorrects over time), with an average of ×2.33. procedure. The purpose of this study was to define a basic
Behav Analysis Practice (2021) 14:679–703 697

Sarah

Fig. 13 Results for Sarah. Note. The top chart shows the raw data for the basic procedure using the Chinese deck. The bottom chart shows the raw data
for the incremental sprints procedure using the Russian deck

SAFMEDS procedure, along with four variations of levels of incorrect responses. These general findings are
SAFMEDS procedures, and to determine the effects of each somewhat difficult to interpret in relation to previous find-
procedural variation on correct and incorrect responses. Taken ings because it is not clear if these procedures are the same
together, the data obtained in this study indicate the basic as or different from those in previous studies. As noted
SAFMEDS procedure, as defined previously, produces previously, many earlier studies of SAFMEDS have vague
low levels of correct responses and higher levels of incor- definitions of the procedures (see Quigley et al., 2017) or
rect responses. The supplemental procedures, in general, include what we defined as variations of SAFMEDS pro-
produced higher levels of correct responses and lower cedures (e.g., additional timings).
698 Behav Analysis Practice (2021) 14:679–703

Lacy

Fig. 14 Results for Lacy. Note. The top chart shows the raw data for the basic procedure using the Chinese deck. The bottom chart shows the raw data for
the incremental sprints procedure using the Russian deck

It should be noted that no additional instruction was of learning or as part of the overall instruction. The rela-
provided on the Russian or Chinese characters other than tion between SAFMEDS and ongoing instruction has not
exposure to the SAFMEDS cards. One might criticize the been made clear in the literature. It is unclear whether
study for this lack of instruction. However, SAFMEDS is these procedures would have been more effective if they
described as both an assessment and an instructional pro- had been used in the context of ongoing, direct instruc-
cedure (e.g., Eshleman, 2000a; Lindsley, 1996). The pre- tion. Future research should evaluate the effects of the
viously published literature is unclear as to whether SAFMEDS procedures in conjunction with other instruc-
SAFMEDS procedures were used strictly for assessment tional opportunities (e.g., lecture, book reading) to
Behav Analysis Practice (2021) 14:679–703 699

Table 3 Level, Celeration, and Bounce Metrics of Correct and Incorrect Responses Across Participants in the Whole-Deck Supplemental SAFMEDS
Procedure

Participant name Condition Language Corrects Incorrects

Level Celeration Bounce Level Celeration Bounce

Karly Basic SAFMEDS Chinese 2.52 x1.0 x4.2 17.8 ÷1.01 x1.3
Whole deck Russian 7.35 x1.27 x2.7 11.71 ÷1.11 x1.9
Terri Basic SAFMEDS Russian 1.27 x1.33 x4.3 23 x1.05 x1.6
Whole deck Chinese 14.91 x1.61 x3.6 6.59 ÷1.42 x2.3
Louise Basic SAFMEDS Chinese 1.85 x1.33 x3.6 20.73 x1.05 x1.3
Whole deck Russian 15.62 x1.63 x2.4 10.54 ÷1.1 x1.5

determine how these additional, non-fluency-based in- Interestingly, in most cases, the variations of the
structional methods impact performance with multiple SAFMEDS procedure produced increases in correct
SAFMEDS procedures. responding and decreases in incorrect responding, suggesting

Table 4 Level, Celeration, and Bounce Metrics of Correct and Incorrect Responses Across Participants in the Incremental Whole-Deck Supplemental
SAFMEDS Procedure

Participant name Condition Language Corrects Incorrects

Level Celeration Bounce Level Celeration Bounce

Jackson Basic SAFMEDS Russian 0.60 x1.09 x2.3 10.76 ÷1.03 x1.6
Incremental whole deck Chinese
Subdeck 1 33.18 x3.68 x1.2 1.82 ÷47.8 x1.2
Subdeck 2 37.51 x2.14 x1.1 1.82 ÷11.3 x2.5
Subdeck 3 21.06 x3.12 x2.5 1.51 ÷2.39 x8.1
Subdeck 4 28.99 x1.06 x1.1 1.26 ÷2.44 x1.7
All decks 21.79 x1.07 x1.8 3.74 ÷1.2 x3.0
Bailey Basic SAFMEDS Chinese 3.02 x1.09 x3.1 33.84 x1.02 x1.4
Incremental whole deck Russian
Subdeck 1 14.59 x3.83 x1.3 6.34 ÷3.0 x2.8
Subdeck 2 18.2 x1.37 x1.6 7.63 ÷1.4 x1.5
Subdeck 3 29.95 x4.13 x1.6 2.88 ÷46.8 x2.0
Subdeck 4 36.65 x1.33 x1.0 1.15 ÷11.3 x4.5
All decksa — — — — — —
Joan Basic SAFMEDS Russian 1.16 x1.18 x3.3 32.93 x1.05 x1.4
Incremental whole deck Chinese
Subdeck 1 31.77 x1.13 x1.4 1.0 x1.0 x1.0
Subdeck 2 32.17 x3.8 x1.3 0.63 x1.0 x2.4
Subdeck 3 34.55 x1.74 x1.5 0.84 ÷1.45 x3.7
Subdeck 4 33.97 x1.16 x1.3 1.0 ÷1.83 x2.0
All decks 27.31 x1.28 x1.8 2.57 ÷1.63 x2.8
Pam Basic SAFMEDS Chinese 2.79 x1.17 x4.5 15.12 x1.03 x1.6
Incremental whole deck Russian
Subdeck 1 23.33 x2.62 x1.4 0.71 ÷2.83 x1.8
Subdeck 2 26.72 x10.2 x1.6 0.79 ÷128 x2.4
Subdeck 3 28.31 x1.18 x1.2 1.26 ÷1.41 x1.1
Subdeck 4 31.66 x1.12 x1.0 0.5 ÷1.0 x1.0
All decks 20.49 x1.52 x1.3 2.74 ÷6.16 x2.9
a
Levels cannot be computed due to only one session with all subdecks combined
700 Behav Analysis Practice (2021) 14:679–703

Table 5 Level, Celeration, and Bounce Metrics of Correct and Incorrect Responses Across Participants in the Sprints Supplemental SAFMEDS
Procedure

Participant name Condition Language Incorrects Corrects

Level Celeration Bounce Level Celeration Bounce

Kit Basic SAFMEDS Russian 1.03 x1.17 x3.3 28.14 x1.03 x1.3
Sprints Chinese 9.37 x1.34 x3.2 17.13 ÷1.08 x1.9
Diane Basic SAFMEDS Chinese 2.46 x1.35 x4.0 31.32 x1.04 x1.3
Sprints Russian 7.90 x1.44 x3.5 31.02 x1.01 x1.4
Myra Basic SAFMEDS Chinese 1.52 x1.12 x4.7 31.35 x1.04 x1.4
Sprints Russian 1.72 x1.27 x3.7 27.55 x1.03 x1.5
Kylie Basic SAFMEDS Russian 2.29 x1.31 x3.1 39.45 x1.08 x1.2
Sprints Chinese 17.77 x1.48 x3.5 32.52 x1.05 x1.2

that they were sufficient to produce at least some learning in procedures was introduced was large, especially when com-
the absence of other, ongoing, explicit, direct instruction. This pared to performance during the basic SAFMEDS procedure.
is in stark contrast to the results for the basic SAFMEDS Of the variations of SAFMEDS procedures, the incremental
procedure. The whole-deck, incremental whole-deck, sprints, whole-deck and incremental sprints procedures appeared to
and incremental sprints procedures all produced an improve- have the most impact on performance. This finding is inter-
ment in correct responding for all participants, except one. esting given Lindsley’s (1996) rationale for his acronym
The change in level when each of the supplemental choice. Specifically, “all,” “fast,” and “minute” were chosen

Table 6 Level, Celeration, and Bounce Metrics of Correct and Incorrect Responses Across Participants in the Incremental Sprints Supplemental
SAFMEDS Procedure

Participant name Condition Language Corrects Incorrects

Level Celeration Bounce Level Celeration Bounce

Nathan Basic SAFMEDS Russian 1.38 x1.14 x3.8 23.74 ÷1.05 x1.8
Incremental sprints Chinese
Subdeck 1 8.55 x3.08 x3.9 13.74 ÷1.8 x1.9
Subdeck 2 11.58 x89.8 x2.5 5.73 ÷320 x3.2
Subdeck 3 13.12 x2.82 x2.5 7.08 ÷2.51 x3.2
Subdeck 4 15.58 ÷1.09 x2.5 6.32 ÷1.33 x3.0
All decks 14.69 x1.61 x1.7 7.49 x1.02 x3.5
Sarah Basic SAFMEDS Chinese 3.01 x1.11 x5.3 20.45 x1.04 x1.6
Incremental sprints Russian
Subdeck 1 15.08 x15 x15 3.98 ÷907 x3.8
Subdeck 2 8.07 x10.6 x10.6 6.79 ÷9.13 x4.5
Subdeck 3 18.54 x2.52 x2.52 4.03 ÷8.36 x6.2
Subdeck 4 15.13 x5.25 x5.25 6.31 ÷21.3 x4.7
All decks 14.4 x1.28 x1.28 12.68 ÷1.0 x2.2
Lacy Basic SAFMEDS Chinese 4.74 x1.1 x3.3 29.42 x1.01 x1.01
Incremental sprints Russian
Subdeck 1 12.32 x3.91 x1.9 17.61 ÷1.84 x1.6
Subdeck 2 18.24 x2.26 x1.4 17.27 ÷2.22 x1.4
Subdeck 3 16.18 x1.46 x1.7 18.96 ÷1.36 x1.3
Subdeck 4 16.92 x2.17 x1.7 16.18 ÷2.13 x1.5
All decks 10.14 x1.05 x1.6 22.17 x1.09 x1.3
Behav Analysis Practice (2021) 14:679–703 701

to indicate limitations of previous flash card methods. “All” study reported having multiple years of experience with a
and “fast” indicated a need to review “all” material at one time second language (though not these languages), were high-
rather than smaller chunks of material (e.g., learning all 60 performing college students (i.e., most had GPAs of 3.0 and
cards instead of 15 at a time). His hypothesis was that doing above), and had used SAFMEDS procedures in their
so would prevent slow, accurate responding with the later coursework prior to enrolling in the study. It is unknown
need to develop fluency. Results of the current study suggest how these factors affected performance. Future studies may
that reviewing the cards in smaller units (i.e., 15 at a time) in a want to include participants who are more naive to
relatively slow manner (i.e., studying them for a 10-min peri- SAFMEDS, are lower performing, and/or have less experi-
od) with subsequent 1-min timings may lead to quick acqui- ence with skills related to those being taught.
sition and steep increases in fluency. This conclusion is based Another potential limitation of the study is that we did not
upon anecdotal data and should be viewed cautiously, counterbalance the order in which participants completed tim-
however. ings with the different decks across phases. That is, we did not
Participants were not required to use the full 10 min allot- control for whether participants completed their Chinese deck
ted for study. Often, participants did not use the full 10 min of or their Russian deck first each day, and we did not control for
study and, furthermore, their study time decreased with each this across the SAFMEDS variations. If participants repeated-
subsequent day. Additionally, participants studied the cards in ly completed the Chinese deck timing after the Russian deck
a “fast” manner during the review period. Specifically, partic- timing, performance could have been higher on the Chinese
ipants often studied the cards in a similar format to the timing deck because the Russian deck served as a “warm-up” (see
procedure—that is, quickly going through the deck, saying Donahoe & Palmer, 2004, for a discussion of priming). If this
cards quickly, and checking their answers. Finally, although phenomenon were observed, it would be especially problem-
we did not program for it, the participants often sued error self- atic after the variations of SAFMEDS procedures were intro-
correction procedures that had been implemented in other duced because the phenomenon might have partially
SAFMEDS studies (e.g., Hughes, Beverly, & Whitehead, accounted for the improved results for the variation strategies.
2007). For example, the participants often set aside the cards To assess this possible limitation, we reviewed the sequence
on which they made errors and then conducted additional of timings for each participant across each phase. Order was
practice on those cards during their study time. This anecdotal classified as balanced (i.e., each deck was practiced first be-
information suggests that perhaps participants were using a tween 40% and 60% of the time) or unbalanced (i.e., either
practice procedure that exemplified the “fast” and “minute” deck was practiced first greater than 60% of opportunities).
components of SAFMEDS. When both decks were exposed to the basic SAFMEDS pro-
This study has its limitations and, thus, results should be cedure, 10 of 14 participants (i.e., Karly, Terry, Louise, Diane,
interpreted with caution. First, participants did not complete Kit, Jackson, Nathan, Lacy, Joan, and Myra) had a balanced
daily timings because they did not practice on weekends. It is order. The performance of these 10 participants was compared
reasonable to assume that participants would have had a to the remaining 4 participants, whose practice was more un-
higher number of corrects and fewer errors if they had had balanced (i.e., one deck more routinely followed another). We
additional learning trials that would have been afforded observed similar performance across all individuals, indicat-
through true daily practice. We did not allow participants to ing no warm-up effects. When the SAFMEDS variations were
take home their decks and do timings on the weekends be- introduced, 9 of the 14 participants (i.e., Louise, Terri, Dianne,
cause we could not be there to conduct those timing sessions. Jackson, Bailey, Nathan, Sarah, and Kylie) had an unbalanced
Furthermore, allowing participants to take decks home could order. In these cases, the deck being practiced with the basic
have added confounds due to additional practice the partici- procedure was typically practiced first, followed by the deck
pants may have done outside of research sessions. being practiced with the SAFMEDS variation. However,
Another limitation was that participants were not allowed comparisons of levels and trends across these participants as
to continue practicing with their decks until rates of perfor- compared to those whose practice was more balanced indicat-
mance stabilized. This limitation is most important for the ed similar performance. Again, this suggested that no warm-
participants who experienced the incremental procedures. up effect was observed. There was one exception, Myra (i.e.,
Best practice probably would have been for participants to crossover never occurred after the introduction of the supple-
practice each subdeck until a specific mastery criterion was mental procedure). Based on this information, it is believed
met. In this study, we provided new decks based on the pas- that the order effects are minimal, if present at all.
sage of time, not performance. If participants had continued Fluency-based instruction has proposed specific outcomes.
practicing their subdecks until high and stable rates of correct These outcomes are retention, endurance, application, and
responding were observed, the decrease in correct responses performance standards (REAPS; Merbitz et al., 2004).
when reintroducing the entire deck might have been less like- Although previous studies have attempted to evaluate the
ly. Another possible limitation is that many participants in this REAPS outcomes of SAFMEDS (e.g., Kim et al., 2001), this
702 Behav Analysis Practice (2021) 14:679–703

study did not. The current study indicates celeration differ- with statistical concepts. European Journal of Behavior Analysis,
10(2), 183–195. https://doi.org/10.1080/15021149.2009.11434321.
ences across procedures, but future studies should identify
Bolich, B., & Sweeney, W. J. (1996). An eleven-year old girl’s use of
whether procedures produce different REAPS outcomes. repeated readings, SAFMEDS, and see/write-think/write practice to
These learning outcomes are the core feature of using develop fluent reading in Hebrew. Journal of Precision Teaching,
fluency-based instruction methods. 14(1), 41–56.
Bower, B., & Orgel, R. (1981). To err is divine. Journal of Precision
In summary, the basic SAFMEDS procedure, as we de-
Teaching, 2(1), 3–12.
fined it, did not appear to facilitate fluency development with Byrnes, M. E., Macfarlane, C. A., Young, K. R., & West, R. P. (1990).
Chinese characters and Russian words in the absence of other Using precision teaching to teach minimum competency test skills.
instruction with college students. However, variations of the Teaching Exceptional Children, 22(4), 58–61. https://doi.org/10.
1177/004005999002200415.
basic SAFMEDS procedure all produced relatively robust in-
Calkin, A. B. (2005). Precision teaching: The standard celeration charts.
creases in correct responses per 1-min timing, as well as de- The Behavior Analyst Today, 6(4), 207–215.
creases in incorrect responses. These findings suggest that CentralReach. (2020). PrecisionX [Computer software]. https://
there is still much to learn about the utility of SAFMEDS precisionx.centralreach.com/
Donahoe, J. W., & Palmer, D. C. (2004). Learning and complex
and the effects of variations of this procedure on the accuracy
behavior. Ledgetop Publishing.
and fluency of student responding. Furthermore, the refine- Doughty, S. S., Chase, P. N., & O’Shields, E. M. (2004). Effects of rate
ment of the SAFMEDS procedure is only the beginning of building on fluent performance: A review and commentary. The
needed SAFMEDS research (Quigley et al., 2017). Further Behavior Analyst, 27(1), 7–23. https://doi.org/10.1007/
replication of these results is warranted before firm conclu- BF03392086.
Eaton, M., & Fox, S. (1983). Using precision teaching to teach precision
sions can be drawn. Future research should focus on further teaching. Journal of Precision Teaching, 3(4), 91–96.
analyses of these and other supplemental procedures to deter- Eshleman, J. W. (1985). Improvement pictures with low celerations: An
mine which procedures lead to quicker fluency development early foray into the use of SAFMEDS. Journal of Precision
and under what conditions these procedures are most effective Teaching, 6(3), 54–63.
Eshleman, J. W. (2000a). Guidelines and considerations for SAFMEDS.
(i.e., in the presence or absence of other explicit instruction). Standard Celeration Chart Topics. http://
Replications of this work with different materials, content standardcelerationcharttopics.pbworks.com/w/page/15573489/
areas, populations, and learning channels are also warranted. SAFMEDS%20on%20the%20Web
Beyond these areas of future research, there is also a need for Eshleman, J. W. (2000b). SAFMEDS issues: “The smudge factor”—
Adventures in stimulus control. Standard Celeration Chart Topics.
researchers to parse out the roles of practice effects and rein- http://standardcelerationcharttopics.pbworks.com/w/page/
forcement rates, if any, with SAFMEDS. Doughty, Chase, 15573488/SAFMEDS%20Issue
and O’Shields (2004) provided an in-depth discussion of the Graf, S. A., & Auman, J. (2005). SAFMEDS: A tool to build fluency. Graf
potential interactions between practice effects, rate of rein- Implements.
Hughes, J. C., Beverly, M., & Whitehead, J. (2007). Using precision
forcement, and fluency-building exercises that have not been teaching to increase the fluency of word reading with problem
thoroughly explored to date. There is still much work to be readers. European Journal of Behavior Analysis, 8(2), 221–238.
done to understand SAFMEDS and all the variations of the https://doi.org/10.1080/15021149.2007.11434284.
general procedure. Kazdin, A. E. (1982). Single-case research designs: Methods for clinical
and applied settings. Oxford University Press.
Kazdin, A. E. (2003). Research design in clinical psychology (4th ed.).
Declarations Allyn and Bacon.
Kim, C., Carr, J. E., & Templeton, A. (2001). Effects of fluency building
Ethical approval All procedures performed in this study involving hu- on performance over “long” durations and in the presence of
man participants were in accordance with the ethical standards of Western distracting social stimulus. Journal of Precision Teaching and
Michigan University (HSIRB Project No. 13-07-27) and with the 1964 Celeration, 17(2), 7–26.
Helsinki declaration and its later amendments or comparable ethical Lindsley, O. R. (1996). Is fluency free-operant response-response
standards. chaining? The Behavior Analyst, 19(2), 211–224. https://doi.org/
10.1007/BF03393165.
Conflict of interest The authors declare that they have no conflict of McDade, C. E., Austin, D. M., & Olander, C. P. (1985). Technological
interest. advances in precision teaching: A comparison between computer-
testing and SAFMEDS. Journal of Precision Teaching, 6(3), 49–53.
McDade, C. E., & Olander, C. P. (1990). SAFMEDS design: A compar-
Informed consent Informed consent was obtained from all individual ison of three protocols. Journal of Precision Teaching, 7(2), 69–73.
participants included in the study.
Meindl, J. N., Ivy, J. W., Miller, N., Neef, N. A., & Williamson, R. L.
(2013). An examination of stimulus control in fluency-based strate-
gies: SAFMEDS and generalization. Journal of Behavioral
Education, 22(3), 229–252. https://doi.org/10.1007/s10864-013-
9172-6.
References Merbitz, C., Vieitez, D., Merbitz, N. H., & Pennypacker, H. S. (2004).
Precision teaching: Foundations and classroom applications. In D. J.
Beverly, M., Hughes, J. C., & Hastings, R. P. (2009). What’s the proba- Moran & R. W. Malott (Eds.), Evidence-based educational methods
bility of that? Using SAFMEDS to increase undergraduate success (pp. 47–62). Elsevier Academic Press.
Behav Analysis Practice (2021) 14:679–703 703

Olander, C. P., Collins, D. L., McArthur, B. L., Watts, R. O., & McDade, Stockwell, F., & Eshleman, J. (2010). A case-study using SAFMEDS to
C. E. (1986). Retention among college students: A comparison of promote fluency with Skinner’s verbal behavior terms. Journal of
traditional versus precision teaching. Journal of Precision Teaching, Precision Teaching and Celeration, 26(1), 33–40.
6(4), 80–82. West, R. P., Young, K. R., & Spooner, F. (1990). Precision teaching: An
Quigley, S. P., Peterson, S. M., Frieder, J. E., & Peck, K. (2017). A introduction. Teaching Exceptional Children, 22(3), 4–9. https://
review of SAFMEDS: Evidence for procedures, outcomes and di- doi.org/10.1177/004005999002200301.
rection for future research. The Behavior Analyst, 41(1), 283–301.
https://doi.org/10.1007/s40614-107-0087-8. Publisher’s Note Springer Nature remains neutral with regard to jurisdic-
tional claims in published maps and institutional affiliations.