12.01.2015 Views

RESEARCH METHOD COHEN ok

RESEARCH METHOD COHEN ok

RESEARCH METHOD COHEN ok

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

228 INTERNET-BASED <strong>RESEARCH</strong> AND COMPUTER USAGE<br />

deliberate non-response (e.g. if they do not wish<br />

to reveal particular information, or if, in fact, the<br />

question does not apply to them, or if they do not<br />

know the answer). Hence the advice of Dillman<br />

et al. (1999)istoavoidthispractice.Oneway<br />

to address this matter is to give respondents the<br />

opportunity to answer an item with ‘prefer not<br />

to answer’ or ‘don’t know’. The point that relates<br />

to this is that it is much easier for participants<br />

in a web-based survey to abandon the survey – a<br />

simple click of a button – so more attention has to<br />

be given to keeping them participating than in a<br />

paper-based survey.<br />

Redline et al. (2002) suggest that branching<br />

instructions (e.g. ‘skip to item 13’, ‘go to item 10’;<br />

‘if ‘‘yes’’ go to item 12, if ‘‘no’’ then continue’)<br />

can create problems in web-based surveys, as<br />

respondents may skip over items and series of<br />

questions that they should have addressed. This<br />

concerns the location of the instruction (e.g. to<br />

the right of the item, underneath the item, to the<br />

right of the answer box). Locating the instruction<br />

too far to the right of the answer box (e.g. more<br />

than nine characters of text to the right) can<br />

mean that it is outside the foveal view (2 degrees)<br />

of the respondent’s vision and, hence, can be<br />

overlo<strong>ok</strong>ed. Further, they report that having a<br />

branching instruction in the same font size and<br />

colour as the rest of the text can result in it<br />

being regarded as unimportant, not least because<br />

respondents frequently expect the completion of<br />

aformtobeeasierthanitactuallyis.Hencethey<br />

advocate making the instruction easier to detect<br />

by locating it within the natural field of vision of<br />

the reader, printing it in a large font to make it<br />

bolder, and using a different colour. They report<br />

that, for the most part, branching instruction errors<br />

occur because they are overlo<strong>ok</strong>ed and respondents<br />

are unaware of them rather than deliberately<br />

disregarding them (Redline et al. 2002: 18).<br />

The researchers also investigated a range of<br />

other variables that impacted on the success of<br />

using branching programmes, and reported the<br />

following:<br />

<br />

The number of words in the question has an<br />

impact on the respondent: the greater the<br />

number of words the less is the likelihood<br />

of correct branching processing by the reader,<br />

as the respondent is too absorbed with the<br />

question rather than with the instructions.<br />

Using large fonts, strategies and verbal<br />

design to draw attention to branching<br />

instructions leads to greater observance of these<br />

instructions.<br />

The number of answer categories can exert<br />

an effect on the respondent: more than seven<br />

categories and the respondent may make errors<br />

and also overlo<strong>ok</strong> branching instructions.<br />

Having to read branching instructions at<br />

the same time as lo<strong>ok</strong>ing at answer<br />

categories results in overlo<strong>ok</strong>ing the branching<br />

instructions.<br />

Locating the branching instruction next to the<br />

final category of a series of answer boxes is<br />

amuchsaferguaranteeofitbeingobserved<br />

than placing it further up a list; this may<br />

mean changing the order of the list of response<br />

categories, so that the final category naturally<br />

leads to the branching instruction.<br />

Branching instructions should be placed where<br />

they are to be used and where they can be seen.<br />

Response-order effects operate in surveys, such<br />

that respondents in a self-administered survey<br />

tend to choose earlier items in a list rather than<br />

later items in a list (the primacy effect), thereby<br />

erroneously acting on branching instructions<br />

that appear with later items in a list.<br />

Questions with alternating branches (i.e. more<br />

than one branch) may be forgotten by the time<br />

they need to be acted upon after respondents<br />

have completed an item.<br />

If every answer has a branch then respondents<br />

may overlo<strong>ok</strong> the instructions for branching as<br />

all the branches appear to be similar.<br />

If respondents are required to write an openended<br />

response this may cause them to<br />

overlo<strong>ok</strong> a branching instruction as they are so<br />

absorbed in composing their own response and<br />

the branching instruction may be out of their<br />

field of vision when writing in their answer.<br />

Items that are located at the bottom of a page<br />

are more likely to elicit a non-response than<br />

items further up a page, hence if branching

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!