1 
Harnessing Youth 
Volunteers as Literacy 
Leaders (HYVALL) 
Endline Student 
Assessment Report
November, 2014
Submitted by Education Development Center, Inc. to the YMCA
Agreement # AID-OAA-F-12-00021 
2 
Table of Contents 
Executive Summary . ..................................................................................................................................... 4
Overview of the HYVALL intervention .......................................................................................................... 4
Overview of the Endline Assessment Instrument .......................................................................................12
Sample Demographics ................................................................................................................................12
Student Results in Reading .........................................................................................................................16
Summary Reading Results ...........................................................................................................................16
Oral reading fluency results ........................................................................................................................18
Reading Comprehension Results ................................................................................................................22
Conclusions and recommendations ............................................................................................................26
References ..................................................................................................................................................27
Appendices ..................................................................................................................................................27
Data Analysis...............................................................................................................................................28
Study Limitations ........................................................................................................................................29
Instrument reliability and validity ...............................................................................................................30
Correspondence between fluency (WCPM) and comprehension ..............................................................32
3 
Table of Figures 
Figure 1. Literacy volunteers’ mean fidelity of implementation of the after school sessions (n=45) .......... 5
Figure 2. Volunteer home visits in the past month (n=358) ......................................................................... 5
Figure 4. Student attendance at afterschool reading session: do you attend the after school reading 
sessions? ....................................................................................................................................................... 6
Figure 3. Student attendance at reading summer camps (n = 358) ............................................................. 6
Figure 5. Number of Reading Activities Students Perform Outside of the Center, per week (n=214) ......... 7
Figure 6. Student’s Home Literacy Activities (n=214, multiple response).................................................... 8
Figure 7. Students Like to Read and Write (n=628) .................................................................................. 8
Figure 8. Students Opinions of their own Reading (n=626)..................................................................... 9
Figure 9. Intervention Students’ Explanations for Why Their Reading Skills Improved (n=198)................. 9
Figure 10. Distribution of number of literate household members (n=114) .............................................10
Figure 11. Do you think members of your community can help you child become a better reader? 
(n=114) ........................................................................................................................................................10
Figure 12. Parent’s Opinion on Advantages of Community Involvement in Their Child’s Literacy 
Development (n=114) ................................................................................................................................. 11
Figure 13. Parents Opinion of their Child’s Reading Level (n=114) ............................................................11
Figure 14. Regional Distribution of Students Tested (n=332) .....................................................................12
Figure 15. Distribution of Students by Level (n=332) .................................................................................13
Figure 16. Distribution of Students, by Sex ................................................................................................. 13
Figure 17. Distribution of Students Tested by Language Spoken (n=332) ..................................................14
Figure 18. Distribution of Texts Read By Intervention (HYVALL) Students, By Level at Endline (N=205) ...15
Figure 19. Distribution of Texts Read By Control Students, By Level at Endline (N=123) ..........................15
Figure 20 Overall of gains in Words Correct Per Minute (WCPM) across baseline and endline ................16
Figure 21. Overall Summary of Gains in Reading Comprehension (out of 6) Across Baseline and Endline
....................................................................................................................................................................17
Figure 22. Change in Zero Scores in Fluency and Comprehension, Between Baseline and Endline ...17
Figure 23. Summary of Gains in Fluency Performance (WCPM) Across Baseline and Endline by Grade 
Level ............................................................................................................................................................19
Figure 24. Summary of Gains in Fluency Performance (WCPM) Across Baseline and Endline by Sex .......20
Figure 25. Percentage of WCPM for CE test ...............................................................................................21
Figure 26. Percentage of WCPM for CM test ..............................................................................................21
Figure 27. Number of Comprehension Questions Answered Correctly After the First and Second Reading 
for the Intervention Group .........................................................................................................................23
Figure 28. Number of Comprehension Questions Answered Correctly After the First and Second Reading 
for the Control group ..................................................................................................................................23
Figure 29. Summary of Gains In Reading Comprehension (Out Of 6) Across Baseline and Endline by Grade 
Level (At Endline) ........................................................................................................................................24
Figure 30. Number of Comprehension Questions Answered Correctly After the Second Reading, By 
Group and Sex .............................................................................................................................................25
Figure 31. Summary of Gains in Reading Comprehension (Out Of 6) Across Baseline and Endline by Sex 26
4 
Executive summary 
Harnessing Youth Volunteers as Literacy Leaders (HYVALL) is a 2 year education program financed by the 
United States Agency for International Development (USAID) under the All Children Reading Grand 
Challenge and implemented by YMCA Senegal with technical assistance from Education Development 
Center, Inc. It aims to improve literacy rates in 6,000 grade 1 to 6 students using an out of school 
reading program led by community volunteers. These volunteers are each assigned a small group of 
students with reading difficulties to mentor and tutor twice weekly after school and during summer 
camps over the course of 1.5 years. The endline assessment conducted in November 2014 allows for an 
evaluation of the program results with respect to student reading performance in comparison to the 
baseline assessment undertaken in March 2013. These assessments were conducted in randomly 
selected "intervention" schools and control schools in the regions of Dakar, Thies, Kaolack, St. Louis and 
Ziguinchor where HYVALL intervenes. It aims to assess whether or not and how much students in the 
HYVALL intervention groups may have progressed since the intervention began in April 2013. The study 
evaluated student reading fluency and comprehension of a text adapted to the student's level 
(independent or instructional-level text). The study is designed as a longitudinal study: students who 
participated in the baseline assessment were tracked wherever possible for this endline assessment. The 
assessment was implemented by Senegalese YMCA staff trained by EDC assessment advisor. This report 
summarizes the results of the endline assessment and compares these results against the results of the 
baseline study. 
Summary results show that students who participated in the HYVALL intervention exhibited larger gains 
in their fluency and reading comprehension from baseline to endline, compared to their control group 
counterparts (p<.001). These results suggest that an intervention like HYVALL that gives students 
opportunities for reading instruction outside school setting (increased time on task), provides regular 
one-on-one tutoring and mentoring and encourages parental involvement in the reading development 
of their children can lead to dramatic student improvement in reading. 
Overview of the HYVALL intervention
To better understand HYVALL intervention results, data on 
the fidelity of implementation as well contextual 
information was collected throughout the project 
implementation. This section of the report presents a 
summary of the fidelity of implementation data including 
volunteer application of the program, student attendance 
to activities and their engagement with home literacy 
activities, and their opinions on reading. Finally, data from 
interviews with parents of intervention students is also 
presented. This data provides a context from which to 
interpret the student reading results. 
5 
Volunteers’ fidelity of implementation. The design of the HYVALL intervention called for “at-risk” 
students to be tutored twice weekly by a local volunteer in reading and writing for a period of 1.5 years. 
Volunteers were provided with leveled reading materials, lesson plan templates for each level and 
scripted activity guides. Depending on the reading levels of their students (emergent, beginning, or 
transitional), volunteers selected activities listed in their lesson plan templates to carry out in their 45-
minute sessions. Volunteers received a four day training in planning for and applying these activities. 
Part of this training involved a shadowing experience where they would accompany the center 
supervisor in a tutoring session. Thereafter, volunteers were assigned a group of students to tutor. The 
median ratio of volunteer to student was 1 to 10 ranging from a minimum of 8 students to a maximum 
of 20 per one volunteer1
. In order to group students by level, short screening placement assessments 
were individually administered consisting of letter reading, high frequency word reading, and passage 
reading. Students were placed in a given reading group according to their scores on this screening. Once 
these groups were determined, volunteers were assigned to groups of students. In cases where many 
students were placed into the same level, centers would assign two volunteers to the same level. For 
example, some centers had two beginning reading groups, one medium group but no advanced group. 
Given this design, how did HYVALL volunteers 
apply these lessons plans? Using an observation 
tool that was based on the lesson plan 
templates, center supervisors conducted 
observations twice a quarter to track fidelity of 
implementation as well as provide feedback to 
those volunteers with regard to the 
implementation issues. Figure 1 shows that
average fidelity of implementation of the reading 
activities was very high. These results show that 
the program was implemented as designed: 
volunteers followed the session plans as 
modeled in their training and as stated in their reading 
activity guide. 
Volunteer home visits. As part of its design, HYVALL also 
aimed to promote a culture of reading and to encourage 
home environments to be more conducive to students’ 
literacy development. To stimulate this, volunteers were 
asked to visit their tutees’ homes once a quarter. At the 
endline assessment, students were asked to report 
whether or not their YMCA volunteer had visited their 
home in the last month. The large majority of students 
 
1
 Number of students per volunteer was determined by student reading levels. Because students were divided into 
leveled reading groups according to their ability levels, in cases where the majority of students were at a certain 
level, volunteers would sometimes have to take into their group more students than was originally planned. 
95.8%
93.3%
96.7%
Emergent
(N=22)
Beginning
(N=12)
Transitional
(N=15)
Figure 1. Literacy volunteers’ mean fidelity of 
implementation of the after school sessions 
(n=45)
Yes
71.5%
No
28.5%
Figure 2. Volunteer home visits in the 
past month (n=358)
6 
(71.5%) reported that a YMCA volunteer had visited their home to talk about literacy with his or her 
parents Figure 2). 
The intention of these visits was to talk to parents about the importance of reading, how their child was 
progressing in reading, and how they as a family could help support their child in his or her literacy 
development. 
Student attendance in project activities. Student participation in the HYVALL intervention is another 
essential element of program implementation. The two main interventions offered to students through 
HYVALL were two 45-minute reading sessions per week for 1.5 years and a month-long summer camp 
offered during the 2014 summer vacation. For the duration of these summer camps, 45-minute reading 
sessions were held every day of the week for one month. For struggling readers, long vacations or 
breaks are shown to be detrimental to their progress (Entwisle et al., 2000). The goal of these camps 
was to ensure that students would start school 
not having lost what they had attained over the 
past school year and to give them an extra boost 
of confidence in reading prior to re-entering 
school. At the endline assessment, tested 
students were asked whether or not they had 
participated in these camps. As Figure 3 shows, 
69% of students reported having attended the 
summer camp. 
At endline, students were also asked whether or 
not they had participated in the weekly after 
school reading session, and if so, how frequently 
they participated. The figures below show that 
90% of sampled students reported regularly 
attending the after-school reading sessions. From those who participated, 86% of students reported 
attending twice weekly, as intended by the intervention. 
Figure 4. Student attendance at afterschool reading session: do you attend the after school reading 
sessions? 
Yes
90.2%
No
9.8%
13.3%
85.8%
once a week twice a week
How many times a week?
Yes
69%
No
31%
Figure 3. Student attendance at reading 
summer camps (n = 358)
7 
These participation rates were confirmed by students’ weekly submission of the “Reading Passport” 
tool. This tool was given to each student participating in 
sessions. The Reading Passport traveled between the 
student’s home and the center once a week. This Reading 
Passport was intended for students to record when they 
engaged in literacy activities outside of the center. The 
Reading Passport was also presented to parents by volunteers 
during home visits as examples of the types of activities they 
could encourage their children to do at home as literacy 
practice. 
From these Reading Passports, volunteers recorded their 
students’ home activities weekly in a summary sheet. Each 
week, students could have checked a total of three boxes if 
they had engaged in reading practice, writing practice, or in 
storytelling. For the purpose of the study, data for a sample of
intervention students was entered for the last 7 weeks of the 
HYVALL intervention. Over the period of 7 weeks, students on average performed 1.65 home literacy 
activities. The figure below shows percent of students who performed one, two or all three activities on 
a weekly basis. More than half of students performed one out of the three activities. Less than a quarter 
of students (21.5%) completed all three activities every week. 
Figure 5. Number of Reading Activities Students Perform Outside of the Center, per week (n=214) 
Out of the three activities students could do at home, the most popular reported activity was reading 
practice (64.2%) followed by writing practice (57.9%) and then storytelling (43.3%). This is shown in the 
figure below. 
57.5%
21.0%
21.5%
1 activity/week
2 activities/week
3 activities/week
Percent of Students
8 
Figure 6. Student’s Home Literacy Activities (n=214, multiple response) 
These results are important as research shows home literacy practice to be just as important for reading 
development as school-based reading (Save the Children, 2013). This is confirmed by this study which 
shows that when pairing student reading scores with their application of home activities, 10.6% of the 
variation in students’ reading comprehension scores could be explained by their engagement in home￾based literacy activities (p=.000, d=2.2, ES=.074). It is therefore suggested that there is significant added 
value in the home-literacy component of this program. 
Student opinions of their literacy. At the endline assessment, students from both the intervention and 
control groups were also asked about their opinions on literacy. First, students were asked if they liked 
to read and to write. Answers to these questions did not differ significantly between intervention and 
control students. 
Figure 7. Students Like to Read and Write (n=628)
Students were then asked to think about whether or not they thought they were good readers. 
Significantly more intervention students indicated that they thought of themselves as good readers in 
comparison to their control counterparts. A larger percentage of intervention students (94%) also stated 
that they had improved in their reading over the last year. 
64.2%
57.9%
43.3%
reading practice
writing practice
storytelling
Control, 95.9%
Control, 96.6%
Intervention, 
98.3%
Intervention, 
98.9%
Do you like to read?
Do you like to
write?
Percent of students who answered "Yes"
9 
Figure 8. Students Opinions of their own Reading (n=626)
How did students think they became better readers? Those intervention students who answered they 
had improved gave several answers to this question. Just over 40% noticed they had simply gotten to be 
more accurate and faster readers. Around 15% said they were better readers because they attended the 
HYVALL after school reading sessions. Around 13% said they improved because they practiced reading 
often, and 11% said they practiced at home. These responses suggest that HYVALL had an impact on 
students’ opinions of their reading through encouraging reading and writing practice outside of school. 
Figure 9. Intervention Students’ Explanations for Why Their Reading Skills Improved (n=198) 
Parents’ opinions of their children’s literacy. During one of the volunteer’s home visits, parents of 
HYVALL students were interviewed to surmise whether or not the community was literate, valued 
literacy, and to gauge what role they thought the community could play in their child’s literacy 
Control, 78.4%
Control, 78.4%
Intervention, 
90.2%
Intervention, 
93.9%
Do you read well?
Has your reading
improved in the last
year?
Percent of students who answered "Yes"
43%
15%
12%
10%
6%
5%
4%
4%
I read fast and well without making errors
I attended HYVALL after
 school reading sessions
I practice reading often
I read and write a lot at home
I went to the HYVALL reading camp
I participate more in class
I have better grades in school
Other
How did you improve your reading skills ? 
10
development. To gauge literacy rates in the family, parents were asked how many people in their 
household could read. Of the 114 parents sampled, 50% reported that three or more people in their 
household could read, while less than 3% reported that no one in the household could read. 
Figure 10. Distribution of number of literate 
household members (n=114)
Figure 11. Do you think members of your community 
can help your child become a better reader? (n=114)
Overall over 90% of parents responded that the community could help a child become a better reader. 
This response suggests a positive reflection of HYVALL’s community component. 
When parents were asked to elaborate on the advantages of having community members involved in 
their child’s literacy development, parents had several responses. The figure below recapitulates 
parents’ explanations. In the context of HYVALL, a community volunteer-led intervention, these 
responses imply that parents recognize how community members can contribute to helping their 
children be successful in school, with language, as well as serve as models for positive attitudes towards 
literacy. 
2.6%
19.3%
24.6%
50.9%
No one
1 person
2 people
3 or more
4.4%
91.2%
3.5%
No Yes Don’t know
11
Figure 12. Parent’s Opinion on Advantages of Community Involvement in Their Child’s Literacy Development 
(n=114) 
Finally, parents were asked to gauge their child’s reading level. Over 65% of parents stated their children 
were good readers while just fewer than 30% claimed their children were not good readers. 
Figure 13. Parents Opinion of their Child’s Reading Level (n=114) 
Interestingly, the number of reported literate family members in the household correlated significantly 
with parents’ responses to this question, with 10% of the variation in parents’ responses being explained 
by the number of literate family members (p=.001, d= 0.66, ES=0.31). That is to say that the fewer 
literate people are in the household, the less parents thought their children were good readers. Though 
students from these less literate households may have indeed started out in the intervention as “poor 
readers”, the data below shows that they dramatically improved in their reading skills. In the context of 
HYVALL which selected “at-risk” students for reading, this is important to bear in mind when looking at
the student reading results. 
90%
90%
80%
80%
70%
70%
70%
70%
Helps them be successful in school
Helps students better understand language
Help to prepare students for school
Helps to improve my child's participation in class
Helps to improve school attendance rates
Provides a good model for what it is to be a good
student
Motivates the student to be well-behaved in
class
Models positive attitudes towards literacy
What are the advantages of having community members involved
1.8%
27.2%
67.5%
3.5%
Can't read Not a good
reader
A good reader Don't know
How would you describe your child as a reader now?
12
Overview of the Endline Assessment Instrument 
The instrument administered at baseline and endline was designed to measure students’ level of oral 
reading fluency (reading accuracy and speed) and reading comprehension at their "independent” or 
“instructional” level. This level is identified as the test is administered. If a student cannot read the first 
grade-level text presented to them with 75% accuracy, he or she will be asked to read a lower level text. 
This exercise is undertaken at the first reading of the text which is limited to 60 seconds. After the first 
reading of the "independent-level" or “instructional level” text is completed, the student is asked a 
series of comprehension questions on the content of the text he/she managed to read without text 
support. Then, the same text is read by the student a second time. The second reading is timed but is 
not limited to a specific time frame. Then the text is left in front of the student and questions are asked 
about the content of the entire text adding an inferential question which requires the student to use the 
information from the text to provide a “plausible” answer. Analysis of each instrument can be found in 
Appendix A. 
Sample Demographics 
The figures below show the distribution of the sampled students by region, level, sex and language(s) 
spoken. The percentage of students in the sample varies greatly by region. Ziguinchor is most 
represented but seems to have experienced a dramatic change in study participants between baseline 
and endline for the control group. Thies is the least represented but maintains equal distribution across 
groups. This is explained by the significant initial selection of students in CM2 (6th grade) at baseline 
who at the endline assessment have moved onto secondary school. 
Figure 14. Regional Distribution of Students Tested (n=332) 
The percentage of students per level also varies demonstrating a strong student participation in CE2 
(grade 3), in comparison with other levels. However, because the assessment allows the student to step 
down to a lower-level text it should be noted that over 10% of intervention CE students and over 50% of 
control CE students who started with the CE-level text actually ended up taking the CP-level text. 
7.2%
6.6%
8.1%
6.0%
10.2%
11.1%
6.3%
11.7%
20.2%
12.3%
Dakar
Thiès
Kaolack
Ziguinchor
St. Louis
Control Intervention
13
Figure 15. Distribution of Students by Level (n=332) 
The distribution by sex remains fairly equal between intervention and control groups. The results 
presented in this report are disaggregated by sex. 
Figure 16. Distribution of Students, by Sex
The distribution by sex and grade level also remains equal between intervention and control groups. 
Table 2. Distribution by Level and Sex 
Grade level Girls Boys
CE1 – 3
rd grade
Number 31 27
% of grade 53.5% 46.5%
CE2 – 4
th grade
Number 57 59
% of grade 49.2% 50.8%
CM1 – 5
th grade
Number 47 39
% of grade 54.7% 45.3%
CM2 – 6
th grade
Number 38 34
% of grade 52.8% 47.2%
5.4%
16.6%
10.8%
5.4%
12.0%
18.4%
15.1%
16.3%
CE1
CE2
CM1
CM2
Boys, 
Interventio
n, 28.6%
Girls, 
Interventio
n, 33.1%
Boys, 
Control, 
19.3%
Girls, 
Control, 
19.0%
Control
Interventio
Control Intervention
14
It is also important to be aware of the language(s) spoken by the students. The language in which one is 
versed in can play a substantial role in student’s learning to read and comprehend a text. At endline, 
almost all of the study students reported speaking Wolof and French, and 100% reported speaking 
another language (mother tongue), as well. Due to the lack of variation in languages spoken, it is not 
possible to identify potential influences of language on students’ gains in reading. 
Figure 17. Distribution of Students Tested by Language Spoken (n=332) 
NB. Many students speak several languages. 
Distribution of independent texts read by students, by grade-level 
The assessment was designed to capture 
students’ oral reading fluency levels when reading 
a text at their "independent" or “instructional”
level. For example, a tested student in CE2 was 
first given the CE text to read. If the student could 
not read more than 75% of the words correctly, 
he or she was asked to read the CP-level text to 
give him or her the opportunity to demonstrate 
his or her real reading fluency capabilities. This is 
important because assessing a student on a text 
that is too difficult limits the student’s ability to 
demonstrate his or her true reading skills. Given 
this evaluation design, it is imperative to consider not only the education level of the student but also 
their established level of independent reading as it pertains to this assessment. For the intervention 
group, almost 15% of CE1 students and 5% of CE2 students were asked to read the CP-level text because 
the CE text was too difficult for them. For the control group students, almost 30% of students in CE1 and 
28% of CE2 students had to go down one level. This shows that a larger proportion of students in CE1 
and CE2 in the intervention group could read a grade-level text in comparison to their control 
counterparts. 
Control, 97%
Control, 96%
Control, 100%
Intervention, 98%
Intervention, 93%
Intervention, 100%
Wolof
French
Other
languages
15
The majority of CM1 and CM2 intervention students managed to read the CM text but about 15% of 
students finished by reading the CE text and 2% read the CP text. A higher proportion of CM1 and CM2 
control students could not read on grade level than their intervention counterparts: 50% of CM2 
students were asked to drop down to a CE-level text and 18% of CM1 students were asked to read a CP￾level text. 
Figure 18. Distribution of Texts Read By Intervention (HYVALL) Students, By Level at Endline (N=205) 
Figure 19. Distribution of Texts Read By Control Students, By Level at Endline (N=123) 
This distribution of levels of text read is important to keep in mind when looking at the results presented 
below because oral reading fluency and comprehension results are judged at students’ determined 
instructional or independent levels.
CP Text,
2%
CP Text, 
15%
CP Text,
5%
CE Text, 
12%
CE Text, 
6%
CE Text, 85%
CE Text, 95%
CM Text, 86%
CM Text, 94%
CE1
CE2
CM1
CM2
Read Below Grade Level
CP Text, 
17%
CP Text, 
6%
CP Text,15%
CP Text, 28%
CE Text,
22%
CE Text, 50%
CE Text, 73%
CE Text, 72%
CM Text, 61%
CM Text, 44%
CE1
CE2
CM1
CM2
Read Below Grade Level
Read Grade Level Text
Read Grade Level Text
16
Student Results in Reading 
Following the above context of the HYVALL intervention, this section presents the reading results 
derived from the endline assessment conducted in November 2014 at the conclusion of the HYVALL 
project. The results demonstrate gains made since the baseline in March 2013 for those students who 
were retained from baseline to endline. 
Summary Reading Results 
Fluency. As shown in the figure below, students’ overall fluency results for both control and 
intervention groups increased significantly (p=.000) between baseline and endline. Intervention 
students were found to read significantly faster and with greater accurately at endline in comparison to 
their control counterparts (p=.000, d=0.85, ES=0.4). Accounting for the high participation rates in 
intervention activities and similar starting level, this difference can be attributed to the HYVALL 
activities. 
 
Figure 20 Overall of gains in Words Correct Per Minute (WCPM) across baseline and endline 
Reading Comprehension. Following the first reading of the passage, students were asked five 
comprehension questions. Overall, students reading comprehension after the first reading for both 
control and intervention groups significantly increased over baseline and endline. However, differences 
gain scores significantly favored the intervention group (p=.000, d=0.84, ES=0.39). For the six 
comprehension questions asked after the second reading of the passage, students’ average reading 
comprehension scores for both control and intervention groups increased significantly (p=.000) between 
baseline and endline. As shown in the figure below, intervention students, on average, were able to 
answer 3.67 more questions correctly (out of six) than they could at baseline. This gain for intervention 
students showed to be significantly greater than their control counterparts endline (p=.000, d=1.01, 
ES=0.45). This suggests that students who participated in 1.5 years of the intervention developed 
reading comprehension skills. In addition, reading comprehension gains for intervention students were 
found to be significantly correlated with their participation in home reading activities. This is confirmed 
by this study which shows that when pairing student reading scores with their application of home 
activities, 10.6% of the variation in students’ reading comprehension scores could be explained by their 
baseline
29.6
baseline
33.4
endline gain
57.23
endline gain
29.04
Intervention group (N=205)
Control group (N=127)
baseline
Difference in 
Gain Scores 
= 28.19
17
engagement in home-based literacy activities (p=.000, d=2.2, ES=.074). It is therefore suggested that 
there is significant added value in the home-literacy component of this program. 
Figure 21. Overall Summary of Gains in Reading Comprehension (out of 6) Across Baseline and Endline 
Zero scores. At baseline, a number of students had "zero" scores in oral reading fluency and / or reading 
comprehension. A "zero" in oral reading fluency means a student could not read a single word correctly 
or did not make an effort to read at all. A "zero" score in reading comprehension is defined by a student 
not answering any questions correctly, a student who did not attempt to respond or a student who 
failed to read the text and, consequently, was not asked any comprehension questions on this text. It is 
important to understand how the zero scores in oral reading fluency and reading comprehension 
between the baseline and midline may have changed. The figure below shows changed in “zero” scores 
between baseline and endline. 
Figure 22. Change in Zero Scores in Fluency and Comprehension, Between Baseline and Endline 
Assessment results showed that the proportion of students with zero scores decreased in both 
intervention and control groups. At baseline, about 5% of students from the control group could not 
read a single word whereas 24% of the intervention group students could not read a word. However, at 
endline, no intervention students had zero scores whereas 1% of control students still had zero scores. 
For reading comprehension, zero scores for intervention students fell from 49% to 0%. This is a notable 
improvement for intervention students as every student showed to be able to answer at least 1 
baseline
2.0
baseline
3.4
endline gain
3.7
endline gain
1.4
Intervention group (N=205)
Control group (N=123)
baseline
6%
24%
1%
0%
16% 6%
49% 0%
WCPM (Control)
WCPM (Intervention)
Reading Comprehension
(Control)
Reading Comprehension 
(Intervention)
Control Baseline Intervention 
Baseline
Difference 
in Gain 
Scores = 
2.3
18
comprehension question correctly. Zero scores in reading comprehension for control students 
decreased by 10% between baseline and endline. 
Oral reading fluency results 
The oral reading fluency portion of the test was divided into two segments. First, students were asked 
to read a passage at their grade-level. For example, students in CE2 were asked to read a 56 word text in 
French for a given period of 60 seconds. If the student could not manage to read 75% of the text 
correctly in that time, the student was then given a test at a grade-level lower (ex. CP) to read in another 
period of 60 seconds. This section begins by presenting the results of students’ performance on the 
timed fluency portion of the assessment across baseline and endline. 
Oral reading fluency results by grade level. Looking at the results disaggregated by grade, student gains 
in words correct per minute (WCPM) were found to be more dramatic in CE1 and CE2 than in CM. 
Intervention students in these grades gained more significantly as compared to their control 
counterparts. The students in these grades started with the intervention program at the CI and CP 
levels. This suggests that intervening earlier can have a more substantial impact on reading 
improvement than for those students who entered the program in later grades. It is also interesting to 
note that students in CE1 from control and intervention groups started with similar mean scores 
whereas in the other grade levels, control students showed to start out as better readers at baseline. 
For these grade levels, however, students from the intervention group not only gained in their reading 
fluency but significantly surpassed their control counterparts at endline. 
19
Figure 23. Summary of Gains in Fluency Performance (WCPM) Across Baseline and Endline by Grade Level 
Oral reading fluency results by sex. Girls were found to make greater gains in terms of accuracy and 
reading speed across baseline and endline for both intervention and control groups. Girls in both 
intervention and control groups also started out with lower mean fluency scores at baseline in 
comparison to boys. However, at endline, the intervention girls showed to gain more than their male 
counterparts but also surpassed boys’ reading fluency rates, on average at endline. On the other hand, 
control girls gained more than the control boys. Still, they showed to have lower mean fluency rates at 
endline than the boys. Despite this, the differences in gains between girls and boys for both 
intervention and control groups are not statistically significant.. 
baseline
7.9
baseline
7.7
baseline
11.7
baseline,
23.5
baseline
41.2
baseline
48.3
baseline
53.5
baseline
59.6
endline gain
62.9
endline gain
46.8
endline gain
69.7
endline gain
32.9
endline gain
54.3
endline gain
23.2
endline gain
41.7
endline gain
11.2
Intervention group (N=40)
Control group (N=18)
Intervention group (N=61)
Control group (N=55)
Intervention group (N=50)
Control group (N=36)
Intervention group (N=54)
Control group (N=18)
baseline
CM1
CM2
CE1
CE2
20
Figure 24. Summary of Gains in Fluency Performance (WCPM) Across Baseline and Endline by Sex 
Oral reading fluency results by test and grade level. In addition to disaggregating by sex and grade 
level, it is also interesting to look at the distribution of students’ abilities to read a grade-level text. This 
allows us to gauge how students fare when judging them by a grade benchmark. The results below 
show the distribution of accuracy in reading a grade-level for those students who read a text at their 
grade level at both baseline and endline. Because no CP-level students were present at the endline, 
results are only presented for CE and CM levels. The first two figures show that approximately 50% of 
CE students from both intervention and control groups couldn’t read between 80 and 100% of the 
words in a grade-level text correctly with approximately one third of students unable to read more than 
40% of a grade-level text correctly. At endline, however, over 80% of CE intervention students were able 
to read 80 to 100% of words correctly in a grade-level text. A higher percentage of control CE students 
at endline were able to read 80 to 100% of the grade-level text passage correctly but a third of students 
could only read up to 40% of the text passage correctly. 
baseline
31.4
baseline
35.7
baseline
27.2
baseline,
23.5
endline gain
52.0
endline gain
28.3
endline gain
61.8
endline gain
29.8
Intervention group (N=95)
Control group (N=64)
Intervention group (N=110)
Control group (N=63)
baseline
BOYS GIRLS
21
Figure 25. Percentage of WCPM for CE test 
At baseline, CM students’ accuracy in reading a grade-level text was largely around 60 to 100% of words 
read correctly for both groups. At endline, the trend line moved further to the right with over 80% of CM 
intervention students reading between 80 to 100% of the words correctly. For CM control students, this 
trend is similar though a smaller proportion of control students could read between 80 and 100% of a 
grade-level text at endline. 
Figure 26. Percentage of WCPM for CM test 
Overall, the above distributions reveal that the grade-level passages may have been too easy for the 
average CE or CM student. If the texts had been moderately more difficult, normal bell-shaped curves 
would have likely emerged. Still, these results provide a picture of how intervention students became 
better equipped to read a text at their grade-level as a result of the intervention. Baseline endline
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
zero < 20% 20 -
40%
40 -
60%
60 -
80%
80 -
100%
Intervention (n=114) 
Baseline
endline
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
zero < 20% 20 -
40%
40 -
60%
60 -
80%
80 -
100%
Control (n=93)
Baseline
endline
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
zero < 20% 20 -
40%
40 -
60%
60 -
80%
80 -
100%
Intervention (n=33) 
Baseline
endline
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
zero < 20% 20 -
40%
40 -
60%
60 -
80%
80 -
100%
Control(n=65)
22
Reading Comprehension Results 
This section presents the distribution of reading comprehension at endline and the results disaggregated 
by grade level and sex across baseline and endline. During the assessment, comprehension questions 
were posed on two occasions: once after the timed first reading cut off at 60 seconds, and the second 
time after a timed but uncapped second reading of the same text. For questions after the first reading, 
the student did not have the text in front of them to refer to while they answered questions, which 
required drawing from memory to respond. In addition, the student only had to answer the questions 
associated with the portion of the text he or she read. For example, if the student was unable to read 
more than 3 lines in 60 seconds, only the questions associated with these three lines were asked. For 
questions after the second reading, the student retained the text in front of them and was free to refer 
to it to answer questions. For the second round of questions, an implicit (inferential) question was 
added to determine whether the student could use the explicit information in the text to respond to the 
question. 
Distribution of reading comprehension results by group. For both groups at endline, students showed 
to be able to answer more comprehension questions correctly at the second reading when they were 
given the opportunity to read the full text in an unlimited amount of time and were able to refer back to 
it to answer questions. This is clearly shown in the figures below. These figures also demonstrate the 
differences in distribution of results in reading comprehension for both intervention and control groups 
at endline. The differences in mean gain scores between intervention and control groups in reading 
comprehension after the first reading are statistically significant, favoring the intervention group 
(p=.000, d=0.68, ES=0.32). This same trend was found in the differences in mean gain scores in reading 
comprehension after the second reading (p=.000, d=1.12, ES=0.49). For the intervention group, the 
distribution of reading comprehension questions answered correctly skews to the right with over 50% of 
students able to answer 5 out of 5 comprehension questions after the first reading and over 70% of 
students able to answer 6 out of 6 comprehension questions after the second reading. For control 
students, the distribution of questions answered correctly after both the first and second readings is 
more evenly distributed with around 30% of students able to answer 5 out of 5 comprehension 
questions after the first reading and just under 50% of students able to answer 6 out of 6 
comprehension questions after the second reading. Overall, these distributions illustrate how 
intervention students had an easier time answering comprehension questions whether from memory or 
using a text support in front of them. This is not surprising given HYVALL reading sessions focused 
heavily on developing students’ language skills, vocabulary, and comprehension strategies of stories 
read. 
23
Figure 27. Number of Comprehension Questions Answered Correctly After the First and Second Reading for 
the Intervention Group 
Figure 28. Number of Comprehension Questions Answered Correctly After the First and Second Reading for 
the Control group 
Reading comprehension results by grade level. Looking across baseline and endline and disaggregating 
by grade-level, it is interesting to note that, like in reading fluency, the largest gains in reading 
comprehension (at the second reading) are also at the CE1 and CE2 grade levels for both groups. 
However, the differences in gains for these grade levels are statistically significant, favoring the 
intervention group (for CE1: p=.001, d=0.96, ES=0.43, for CE2: p=.000, d=1.4, ES=0.57). This suggests 
that HYVALL not only served to improve reading fluency for the early learners but also was 
simultaneously successful in building their reading comprehension. Gains were also significant for 
intervention and control students in the older grades except for CM2 control students who showed no 
0%
10%
20%
30%
40%
50%
60%
70%
80%
0 1 2 3 4 5 6
Percentage of students
Number of questions answered correctly
After the first reading
After the second reading
0%
10%
20%
30%
40%
50%
60%
0 1 2 3 4 5 6
Percentage of students
Number of questions answered correctly
After the first reading
After the second reading
24
gains across baseline and endline. This can be explained by their high reading comprehension scores at 
baseline. Still, the CM intervention group surpassed their average reading comprehension scores at 
endline. This same trend also recurs across all grade levels suggesting that despite weaknesses in 
reading comprehension at baseline, the intervention group was able to catch up and surpass their 
control counterparts after 1.5 years of intervention. 
Figure 29. Summary of Gains In Reading Comprehension (Out Of 6) Across Baseline and Endline by Grade 
Level (At Endline) 
baseline 
5.0
baseline
0.6
baseline
1.1
baseline
1.5
baseline
3.5
baseline
2.2
baseline
3.7
baseline
3.3
endline loss,
-0.3
endline gain, 5.1
endline gain, 3.5
endline gain, 4.1
endline gain, 1.4
endline gain, 3.4
endline gain, 1.1
endline gain, 2.4
Intervention group (N=40)
Control group (N=15)
Intervention group (N=61)
Control group (N=54)
Intervention group (N=50)
Control group (N=36)
Intervention group (N=54)
Control group (N=18)
baseline
CM1
CM2
CE1
CE2
25
Reading comprehension results by sex. The figures below show the distributions of reading 
comprehension scores for boys and girls at endline. For the intervention group, the results show little 
difference between the performance of girls and boys in reading comprehension. For the control group, 
differences between boys and girls were more apparent: about 50% of girls were able to answer 6 
comprehension questions when only 45% of boys were able to answer 6 questions. Overall in the 
control group, girls tend to outscore boys. 
Figure 30. Number of Comprehension Questions Answered Correctly After the Second Reading, By Group 
and Sex 
Intervention (n=356) Control (n=261)
Looking across baseline and endline, girls gained more in reading comprehension than boys, in both 
intervention and control groups. Comparing the gains made by intervention and control girls, 
intervention girls were also found to gain significantly more (p=.000, d=1.08, ES=0.48). Intervention girls 
also outscored the control boys at endline. 
Intervention boys also gained significantly more than their control boy counterparts, surpassing the 
control boys at endline (p=.000, d=1.07, ES=0.47). However, differences between boys’ and girls’ gains in 
the intervention group were not significant. These results suggest that though the HYVALL intervention 
likely contributed to strong gains in reading comprehension; it did not seem to contribute to 
differentiated gains for boys and girls. This makes sense when considering participation rates in HYVALL 
activities was equally high for both girls and boys alike. 
0.0%
1.8%
0.0%
3.0%
3.6%
17.6%
73.9%
0.0%
0.5%
0.5%
3.7%
4.2%
19.9%
71.2%
Zero
One
Two
Three
Four
Five
Six
Boys
Girls
3.9%
2.3%
5.4%
7.8%
17.1%
17.8%
45.7%
0.8%
3.8%
3.0%
8.3%
13.6%
20.5%
50.0%
Zero
One
Two
Thre
e
Four
Five
Six Boys
Girls
26
Figure 31. Summary of Gains in Reading Comprehension (Out Of 6) Across Baseline and Endline by Sex 
Conclusions and recommendations 
The results of the analysis of reading gains between the baseline and endline showed that the HYVALL 
reading intervention was effective in improving oral reading fluency and comprehension for the 
students they served, regardless of gender and grade level. Though the effect was greater for younger 
students than older students, older students were still found to greatly benefit from the intervention. 
The program effect can be confirmed by the strong implementation and attendance data which 
indicates high fidelity of program implementation. The success of the intervention with respect to 
increasing reading gains is likely due to the increased exposure to literacy activities it offered, as well as 
the close and regular mentorship it provided its beneficiary students. In addition to the activities it 
provided outside of school, the program also promoted student practice at home allowing students to 
take home writing samples they had produced in the centers, to practice reading and writing further. 
This not only served to continue supporting students’ literacy development beyond the centers but also 
linked family members to students’ literacy development, contributing to the establishment of a culture 
of reading. In future design of this program in Senegal or other contexts, it is strongly recommended 
that a similar design be considered as it has potential for scalability without losing quality. However, the 
success of this model depends on the commitment of community volunteers. To foster this 
commitment, it is recommended to implement an incentives system that formally recognizes the 
contribution of the volunteers and grant them credits that could be allocated for university, 
employment or for the civil service. In ensuring volunteers engagement and in providing them with the 
necessary support and quality materials (guide, books) a program like HYVALL promises significant and 
positive results for student literacy gains and school readiness. 
baseline
2.1
baseline
3.5
baseline
1.9
baseline,
3.4
endline gain
3.5
endline gain
1.1
endline gain
3.8
endline gain
1.6
Intervention group (N=95)
Control group (N=61)
Intervention group (N=110)
Control group (N=62)
baseline
BOYS GIRLS
27
References 
Entwisle, D., Alexander, K., & Olson, L. 2000. “Summer Learning and Home Environment.” In R. 
Kahlenberg Ed., A Notion at Risk: Preserving Public Education as an Engine for Social Mobility pp. 9-30. 
New York: Century Foundation Press. 
Save the Children (2013). Beyond School Walls: A Boost for Readers. Found at: 
http://www.savethechildren.org/atf/cf/%7B9def2ebe-10ae-432c-9bd0-
df91d2eba74a%7D/BEYOND_SCHOOL_WALLS_LITERACY_BOOST_2013.PDF
Appendices 
Appendix A: Methodology 
The reading assessments were designed to provide a snapshot of student’s reading capabilities in 
fluency and reading comprehension at their instructional or independent level. This assessment took an 
average of 6 minutes to administer and included the following sub-tests: 
- First reading of a passage. Students ‘reading accuracy and automaticity in reading a passage 
aloud was assessed. The subtest was timed and capped at 60 seconds, allowing the student to 
read until the 60 second mark. This yielded a score of correct words per minute and served to 
determine students’ instructional or independent level. If students made 75% of more errors in 
reading the text or could not get through 75% of the text in a 1-minute period, they were asked 
to restart the first reading of a passage at a grade-level below. Below is the breakdown of 
number of words per passage for each assessment: 
Assessment level Number of words in 
passage
CP (grade 2) 34 words
CE (grade 4) 56 words
CM (grade 6) 75 words
- Reading comprehension questions (after first reading). Students’ capacity for reading 
comprehension was assessed based on their first timed reading of the passage. A total of five 
literal questions were asked. Students could not refer back to the text to help them answer 
questions. 
- Second reading of a passage. Students ‘reading accuracy and automaticity in reading a passage 
aloud was assessed. The passage is the same “instructional” or “independent” level identified 
and read during the first reading. The subtest was timed but not capped at 60 seconds, allowing 
the student to read until the end. This yielded a score of correct words per minute. At the 
second reading, students’ “instructional” or “independent” level is already determined so 
students who struggle with this reading are not dropped down to a lower level. 
- Reading comprehension questions (after second reading). Students’ capacity for reading 
comprehension was assessed based on their first timed reading of the passage. A total of five 
literal questions and a sixth inferential question were asked. Students were allowed to refer 
back to the text to help them answer questions. 
28
A group of administrators drawn from the HYVALL team from the five regions were trained on test 
administration following a specific protocol. The test was piloted to assess and equate three different 
texts with a randomly selected group of CP, CE, and CM classes in Dakar. 40 students from each grade 
level (120 total) participated in the pilot. Following this pilot, texts that provided statistically close 
fluency and comprehension results were selected to use at baseline, midline, and endline. Because 120 
students were insufficient to provide accurate psychometrics scores, the baseline and endline texts for 
each grade level were the same. This was deemed to have no impact on results given students who had 
started with a text at baseline had moved up to another grade level at endline. 
Sampling: At baseline, in March 2013, EDC randomly selected control and intervention schools. From 
these schools, test administrators were instructed in how to randomly select 5 girls and 5 boys from 
each grade-level in these schools to retain for the assessment for a total of 570 students. For midline 
and endline assessments, those same students who were still present at their schools were assessed. 
For those who had left the school, test administrators were trained to randomly select replacement 
students at this same grade-level and sex. 
Data entry: At baseline, assessments were conducted using a paper-based tool and thereafter, results 
were entered into a data shell in Excel. At midline and endline, data from assessments was collected live 
and electronically using Survey To Go. This was to in order to minimize data entry errors and to expedite 
data collection and entry process. 
Missing data analysis 
Missing data analysis of those students who dropped out of the sample across baseline and endline 
shows that the majority of students who left the sample were in grade 5 and 6 at the baseline which 
places them in secondary school at the endline. As these assessments were carried out in the primary 
school setting, these students would have been eliminated from the sample. Apart from these cases, 
twice as many control students left the sample in comparison to intervention students. T-test 
comparisons of mean scores of baseline show statistically significant differences between the fluency 
(WCPM) of intervention students who left and those control students who left (favoring intervention 
students) (p=.012). Furthermore, in comparing those students who left the sample to those who 
remained, statistically significant differences in baseline comprehension scores (p=.022) of those who 
were retained were found over those who dropped out. These are interesting to note because it 
suggests that those with weaker scores dropped out for certain reasons. Further study is needed to 
determine the reasons for these dropouts. 
Data Analysis 
All collected data were cleaned by EDC M&E staff and analyzed using standard statistical techniques, as 
needed for different analytical purposes. The results were disaggregated by sex and grade-level, as 
appropriate. Central tendency analysis (e.g. mean, frequencies, median) conducted for continuous 
demographic variables. Comparison of means statistical tests (paired and independent samples t-test) 
were conducted to estimate differences between groups such as grade level, type of participant (control 
29
vs. intervention) and sex, where appropriate. Bivariate statistical analyses (e.g., correlations) were 
conducted to examine the relationship between different variables. 
Study Limitations 
The differences in mode of administration between the baseline and the endline assessment potentially 
complicated the collection and entry of data, especially in the baseline study. For the baseline study, 
data was collected using paper-based instruments and was then entered into an Excel shell. This may 
have led to a certain percentage of data entry error even despite 100% verification by two agents and 
again by an EDC agent prior to the endline assessment. At endline, instruments were administered using 
a tablet programmed with Survey-To-Go. This allowed for live data collection and included safeguards to 
minimized administration error (i.e. timing and calculations). Due to differences in administration, data 
quality may not be similar across baseline and endline. 
At baseline, it was discovered that the control group had significantly better fluency and comprehension 
scores in comparison to the intervention group scores. Although the samples were selected at random 
(the control group was selected from comparable schools to the intervention schools in region). 
However, students selected from the control schools were randomly selected without considering their 
reading level. On the other hand, the students selected to participate in the HYVALL program and hence, 
students from the “intervention” were students who had been referred to the program due to their 
weaknesses in reading. This is because the goal of the intervention was to target students with reading 
difficulties. Hence, from the get-go students in the control group were already well ahead in reading 
compared to their “intervention” counterparts. This indicates that the samples are not really 
comparable. 
Fidelity of implementation data was also incomplete for some of the variables. For example, no data was 
provided for Thies for the outside of center activities. The reason for this that volunteers in this region 
did not keep track of students’ reading passports. 
30
Instrument reliability and validity 
It is not possible to provide reliability results for this instrument given the number of sub-tests on the 
test is limited to two per level (fluency and comprehension). However, it is possible to establish the 
concurrent validity of the fluency scores between the first and second reading scores and between 
fluency and comprehension scores. Overall, bivariate correlations between these scores were found 
statistically significant for all instrument levels. That is to say, a linear regression analysis between the 
fluency scores for the first and the second reading show that test scores are significantly and positively 
correlated. Fluency and comprehension scores provide similar results. This is shown in the tables below. 
Concurrent validity: correlations for CP test instrument at endline
Correlations
CP_COMP_Tota
l_correct_1st_re
ading_endline
WCPM_1st_read
ing_CP_endline
CP_COMP_Tota
l_correct_2nd_re
ading_endline
WCPM_2nd_rea
ding_CP_endline
CP_COMP_Total_correct_1st
_reading_endline
Pearson Correlation 1 .773** .815** .764**
Sig. (2-tailed) .000 .000 .000
N 61 61 60 60
WCPM_1st_reading_CP_end
line
Pearson Correlation .773** 1 .607** .712**
Sig. (2-tailed) .000 .000 .000
N 61 68 60 60
CP_COMP_Total_correct_2n
d_reading_endline
Pearson Correlation .815** .607** 1 .613**
Sig. (2-tailed) .000 .000 .000
N 60 60 60 60
WCPM_2nd_reading_CP_en
dline
Pearson Correlation .764** .712** .613** 1
Sig. (2-tailed) .000 .000 .000
N 60 60 60 60
**. Correlation is significant at the 0.01 level (2-tailed).
31
Concurrent validity: correlations for CE test instrument at endline
Correlations
CE_COMP_Tota
l_correct_1st_re
ading_endline
WCPM_1st_read
ing_CE_endline
CE_COMP_Tota
l_correct_2nd_re
ading_endline
WCPM_2nd_rea
ding_CE_endline
CE_COMP_Total_correct_1st
_reading_endline
Pearson Correlation 1 .633** .644** .353**
Sig. (2-tailed) .000 .000 .000
N 309 309 260 260
WCPM_1st_reading_CE_end
line
Pearson Correlation .633** 1 .258** .728**
Sig. (2-tailed) .000 .000 .000
N 309 328 260 260
CE_COMP_Total_correct_2n
d_reading_endline
Pearson Correlation .644** .258** 1 .282**
Sig. (2-tailed) .000 .000 .000
N 260 260 260 260
WCPM_2nd_reading_CE_en
dline
Pearson Correlation .353** .728** .282** 1
Sig. (2-tailed) .000 .000 .000
N 260 260 260 260
**. Correlation is significant at the 0.01 level (2-tailed).
Concurrent validity: correlations for CM test instrument at endline
Correlations
WCPM_1st_readin
g_CM_endline
CM_COMP_Tot
al_correct_1st_r
eading_endline
WCPM_2nd_readi
ng_CM_endline
CM_COMP_Tot
al_correct_2nd_r
eading_endline
WCPM_1st_reading_C
M_endline
Pearson Correlation 1 .669** .692** .226**
Sig. (2-tailed) .000 .000 .000
N 393 383 299 299
CM_COMP_Total_corr
ect_1st_reading_endli
ne
Pearson Correlation .669** 1 .352** .680**
Sig. (2-tailed) .000 .000 .000
N 383 383 299 299
WCPM_2nd_reading_
CM_endline
Pearson Correlation .692** .352** 1 .190**
Sig. (2-tailed) .000 .000 .001
N 299 299 299 299
CM_COMP_Total_corr
ect_2nd_reading_endli
ne
Pearson Correlation .226** .680** .190** 1
Sig. (2-tailed) .000 .000 .001
N 299 299 299 299
**. Correlation is significant at the 0.01 level (2-tailed).
32
Correspondence between fluency (WCPM) and comprehension 
The figure below provides an illustrative snapshot of the correspondence between fluency in 
comprehension. For the purpose of setting benchmarks, these numbers are useful because they can 
serve to guide stakeholders in the identification of thresholds for fluency based on real data. 
0
20
40
60
80
100
120
140
160
180
200
0 1 2 3 4 5 6
47.9
31.5
46.96 56.89
74.17
86.5
94.1
Number of questions correctly answered
Words correctly read per minute (WCPM)
33
Table 3. Summary results from intervention groups (HYVALL) across baseline and endline 
Table 4. Summary results from control groups across baseline and endline 
Description N Mean St. Dev P-value Effect size
Percentage of words read correctly (accuracy)
CE level (percentage)
Baseline 95 26.14% 29.8%
Endline 101 93.22% 13.09% .000 0.9
CM level (percentage)
Baseline 90 60.55% 39.89%
Endline 104 95.57% 7.38 .000 0.66
Number of words read correctly (WCPM) 
CE level 
Baseline 101 10.22 16.87
Endline 101 77.21 27.92 .000 0.9
CM level 
Baseline 115 47.55 35.76
Endline 104 95.29 26.25 .000 0.74
Comprehension questions answered correctly after the first reading (percentage) 
CE level (out of 5)
Baseline 101 20.2% 29.4%
Endline 101 88.4% 19.6% .000 0.89
CM level (out of 5)
Baseline 104 39.4% 38%
Endline 104 82.6% 20.1% .000 0.71
Comprehension questions answered correctly after the second reading (percentage) 
CE level (out of 6)
Baseline 101 19.1% 30.8%
Endline 101 94.1% 12.1% .000 0.92
CM level (out of 6)
Baseline 104 46.6% 40.8%
Endline 104 94.2% 11.3% .000 0.75
Description N Mean St. Dev P-value Effect size
Percentage of words read correctly (accuracy)
CE level (percentage)
Baseline 73 47.64% 35.09%
Endline 73 80.4% 28.48% .000 0.7
CM level (percentage)
Baseline 54 71.23% 31.15%
Endline 54 87.92% 17.18% .000 0.57
34
NB. Grade levels represent student level at endline 
Number of words read correctly (WCPM) 
CE level 
Baseline 73 19.62 18.23
Endline 73 55.95 31.05 .000 0.79
CM level 
Baseline 54 52.1 33.15
Endline 54 71.27 27.57 .000 0.65
Comprehension questions answered correctly after the first reading (percentage) 
CE level (out of 5)
Baseline 69 35.4% 30.2%
Endline 69 72% 30.9% .000 0.71
CM level (out of 5)
Baseline 54 56.6% 31.2%
Endline 54 66.2% 30.04% .035 0.28
Comprehension questions answered correctly after the second reading (percentage) 
CE level (out of 6)
Baseline 69 50.6% 34.1%
Endline 69 80.2% 28.2% .000 0.65
CM level (out of 6)
Baseline 54 69.1% 32.3%
Endline 54 79.6% 23.5% .041 0.27