You are on page 1of 15

Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 1

The placement of cross-referencing material


within a home reference manual

Summary

This user testing report accompanies a home reference manual on

falconry, designed by the author. A feature of the manuals’ design is its

alternative book layout compared with conventional methods found in other

falconry manuals.

The author tests 12 participants, with little or no knowledge of falconry,

by recording the time taken to find cross reference page numbers in two

spread designs; spread A having cross reference under the body text, and

spread B placing them on photographs.

Following a log (base 10) transformation to normalise the results, a t-

test concluded participants were significantly faster at cross referencing in

spread A than in spread B (one-tailed matched pairs t-test; t11= -2.00,

p=0.035).

Test objectives

To enable the user to navigate a falconry book with a new access

structure successfully, it is imperative that a system exists to aid them to find

relevant links and sections quickly and efficiently.

This test aims to determine the effectiveness of two different

navigation methods within a single spread of the book, and calculate if

different design approaches show a significant difference in the time it takes

the user to find a page reference. The results from this test will be useful, not

only in the case of the falconry manual, but any design requiring complex

navigation through its content.


Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 2
 

Test method

Design approach

The section in the book describing raptor (birds of prey) flights and

hunting techniques has a number of links to another section describing

specific falconry tasks. The reader is required to navigate the book using a

cross referencing system. Two design variants were tested; spread A refers to a

design where cross references are located under the body text, whereas spread

B locates the cross references in the photos. The cross references in both tests

share similar design characteristics (type colour, typeface and within a

coloured banner). Appendix A contains the two design approaches seen in the

test spreads.

Procedure

Two spreads of the same book section were printed at actual size, one

using spread A references and the other using spread B. The author carried out

the test with each participant within a quiet room with few distractions,

reading out the statement below before commencement:

 
  “On the page in front of you there are some references to specific
  tasks involved in flying a bird of prey. In a moment I would like you
    to turn over the piece of paper and tell me what page you would
  find an explanation on hooding/flushing game [choose]. I shall be
timing your response but please be aware, this is not a test of your
  ability, I am testing the design effectiveness of this page. I have
  two different pages to try”.
   
Initial participant dialogue

Each participant was asked to find a reference to either ‘hooding’ or

‘flushing game’ in each spread; to reduce the impact of learning they were not

asked to search for the same reference in both spreads. To obviate the
Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 3
 

potential effects of practice and task sequence, half of the participants were

tested on spreads A then B, and the other half B then A (to further reduce any

influence of short-term memory, the reference page numbers for hooding and

flushing game were different on each test spread). Details of participant

questioning can be seen in Appendix B.

The variable this usability test is investigating is the time taken to find

a page reference. It was measured by the author recording the time taken from

turning the spread, to the participant saying out-loud the page number. Once

the test had been timed, each participant was asked to give their opinion on

the spread design, with the author asking which one they preferred and why.

Thinking out-loud after the test, as apposed to during the test, did not distract

the participant during the timed task.

Participants

There were four combinations in which the variables could be asked

(see table below), i.e. it would take four participants to achieve a complete

range, so asking 12 participants would repeat all variations three times. It was

decided that 12 would be sufficient to give a large enough data-set, and this

method of counterbalancing would make sure the order varied and was

balanced. The results were focussed on the time taken to find page references;

results for hooding and flushing game were not compared as this was not the

variable of interest in this test.

Participant Test sequence

1 Hooding on Spread A Flushing game on Spread B

2 Hooding on Spread B Flushing game on Spread A

3 Flushing game on Spread B Hooding on Spread A

4 Flushing game on Spread A Hooding on Spread B

Counterbalanced test method


Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 4
 

Participants were not from a design background and had not been

involved in the design of the test artwork. None were practicing falconers and

there was a mix of gender, age, and country of origin, but all could speak

English fluently.

Statistics

Statistical analysis of the quantitative results was required to ascertain

if the spread design had a significant effect on page referencing times. A t-test

is a parametric test, which requires data to be normally distributed. A

normality test revealed that the results were normally distributed in spread A

(Anderson-Darling normality test, p=0.61), but not normally distributed in

spread B (Anderson-Darling normality test, p<0.005). The results of tests A

and B were normalised using a log (base 10) transformation (Anderson-

Darling normality test; test A, p=0.27, test B, p=0.53). Probability plots of the

normality tests can be seen in Appendix C.

Hypotheses

The test hypothesis for this investigation predicts that the time

participants take to find references when placed under the text, will be quicker

than when references are placed in the photographs. The reasoning for this

hypothesis is due to the participants little understanding of falconry. It would

be expected that a newcomer to the sport would first look at text references

rather than explanatory photos, as they would have little knowledge of what

image relates to the falconry task. The null hypothesis is that the time to find

references under the text will not be quicker than when the reference is placed

in the photographs.
Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 5
 

Limitations

There are a number of constraints in usability testing, namely

reliability and validity (Wenger and Spyridakis, 1989). To increase reliability

this test would need to use more participants to enable a better understanding

of the population and increase the chance of achieving the same results in

multiple tests, but this is made more difficult due to humans being

complicated and subject to a large number of external and internal stimuli.

Validity of the results could be difficult as the findings may actually occur from

a variable other than the placement of the text references.


Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 6
 

Results

Quantitative

The table below illustrates the times taken to find page references in

spreads A and B by 12 participants. Participant 5 may have been a contributing

factor to why spread B results did not have a normal distribution, however, the

log (base 10) transformation solved this issue as mentioned previously in the

test method.

Spread A –
Spread B –
references under Difference
Participant references in
text (time in (seconds)
photos (seconds)
seconds)

1 7.41 4.00 +3.41

2 15.81 39.41 -23.60

3 2.85 23.28 -20.43

4 1.86 11.60 -9.74

5 10.62 120.22 -109.60

6 6.35 3.37 +2.98

7 1.92 10.41 -8.49

8 1.38 6.79 -5.41

9 5.00 2.35 +2.65

10 8.97 3.05 +5.92

11 7.23 7.60 -0.37

12 11.53 40.35 -28.82

Mean 6.74 22.70 -15.96

Standard
4.47 33.55 31.65
deviation

Standard error of
1.29 9.68 9.14
the mean

Time taken for participants to find page references in spreads A and B


Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 7
 

Plotting the results in the diagram below illustrates, on average, the

sample used in this test was 15.96 seconds faster in finding page references in

spread A, with no overlap of the standard error of the mean.

Plotting the standard error of the mean for spreads A and B

Once the data had been normalised it could be analysed using a

parametric test. A matched pairs, one-tailed (due to the use of a directional

hypothesis) t-test was performed, producing the following output:

Paired T-Test and CI: logtime1, logtime2

Paired T for logtime1 - logtime2

N Mean StDev SE Mean


logtime1 12 0.718 0.350 0.101
logtime2 12 1.036 0.531 0.153
Difference 12 -0.318 0.549 0.158

95% upper bound for mean difference: -0.033


T-Test of mean difference = 0 (vs < 0): T-Value = -2.00 P-Value = 0.035
 

Results of a t-test from normalised data using Minitab v.15


Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 8
 

Reviewing the t-test it is possible to conclude participants were

significantly faster in spread A than in spread B (one-tailed matched pairs t-

test; t11= -2.00, p=0.035). Due to this result, the null hypothesis can be

rejected.

Qualitative

A full description of the participant opinions can be seen in Appendix

B. 11 of the 12 participants tested preferred spread A, with page references

under the text. 3 of the 11 who preferred spread A, stated that spread B was not

the convention they were used to.

Influence on final design

The final spreads can be seen in Appendix D. Due to the quantitative

results of this test showing that page referencing placed under the body text is

significantly quicker to read, the falconry manual has been designed using this

method. The final designs align the top of each process step across the spread,

with page referencing kept in standard sized, and consistently coloured, boxes.

Consistency is a key element to the design and was mentioned in the

qualitative results of participant feedback. It is presumed this positioning will

enable the reader (with little understanding of falconry processes) to access

the cross referenced section efficiently and quickly.

Future designs shall presume that inexperienced readers, that is to say

readers with no knowledge of the subject matter, may find references placed

next to body text quicker than if placed on photo elements of the design.
Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 9
 

Acknowledgements

The author sought advice on statistical testing from Patricia Cremona (MSc.

Wildlife Management and Conservation, University of Reading 2010), and

used her licensed copy of Minitab v.15.

References

Null hypothesis. URL: http://en.wikipedia.org/wiki/Null_hypothesis


[13.03.2010]

t-Test for the Significance of the Difference between the Means of Two
Correlated Samples. URL: http://faculty.vassar.edu/lowry/ch12pt1.html
[10.03.2010]

Wenger, M.J., & Spyridakis, J.H. (1989). The relevance of reliability and
validity to usability testing. IEEE Transactions on Professional
Communication, 32(4), 265-271)
Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 10
 

Appendices

Appendix A: Test spreads

Spread A: Page references placed under body text (upper is the left page of the
spread, lower is the right)
Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 11
 

Spread B: Page references placed in photographs (upper is the left page of the
spread, lower is the right)
Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 12
 

Appendix B: Participant results

Participant Test Participant comments

1 1A, 2B Once I’ve learnt the style it’s a quick jump. I preferred
the first one (spread A)

2 1B, 2A I prefer spread A, and like the [reference] boxes being


bigger

3 2B, 1A Spread A is clearer

4 2A, 1B I like spread A where the reference is under what I read

5 1A, 2B Spread A is better as I’m not used to references on


photos. Below the photo would be better

6 1B, 2A Reading through you see spread A references better,


but references in the pictures may be better for
skimming

7 2B, 1A A is easier, I look at the text first. Either way, once you
know it’s in a tan box it’s quick

8 2A, 1B They are very similar, hardly any difference, but I prefer
references near the text

9 1A, 2B I like the colour difference and went through the


numbers first. B is against the convention so A is better

10 1B, 2A I prefer A, the references stand out more than when


they are placed in the pictures

11 2B, 1A Prefer A; I’m used to referencing in the text

12 2A, 1B Prefer the design of B as A seems too condensed. I like


the flow of B

Participant testing and qualitative feedback (1=hooding, 2=flushing game)


Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 13
 

Appendix C: Probability plots of normality tests

Spread A normality test (non-transformed)

Spread B normality test (non-transformed)


Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 14
 

Spread A normality test (Log-transformed)

 
 
Spread B normality test (Log-transformed)

 
 
Oliver Tomlinson – Usability testing: self-directed project (Spring term 2010) 15
 

Appendix D: Final spread design (post usability testing)

 
 
Final spread design (upper is the left page of the spread, lower is the right)
 

You might also like