It means that your factor 1 may be interpreted as positively related with items 1, 2 and 3 whereas factor 2 is inversely related with items 4, 5 and 6.

Very odd sounding, Michael.  It’s up to you which way you frame each factor (e.g., in the Big 5 model some researchers call a factor "emotional stability" and others call the same thing “neuroticism” and reverse the scoring direction).  Just some clarifying questions: How did you assign the original direction/meaning to your factor?  How did you decide it was backward?  Normally the items with strong positive loadings on the factor are scored as original and those with strong negative loadings would be reverse scored.  Did you double-check the way you coded your items into numbers?  For example, if you have a Likert scale and, for example, you put "strongly disagree" as 7 to "strongly agree" as 1, then you might be intuitively flipping your interpretation of the factor around (i.e., because the factor would be defined by strongly disagreeing with the positively loaded items and strongly agreeing with the reverse score items).  Best with finishing your study. ~ Kevin

You should be cautious about any reverse-scored items in that factor. Additionally, running an internal consistency test could identify any problematic items in that component.

Thanks,

-----------

My recent published work:

Conference Paper Constructing and validating a multi-dimensional measure for ...

Conference Paper Exploring the effects of wartime perceptions on attitudes to...

Article Relating Patient Satisfaction to Nurses’ Job Satisfaction, J...

Article Linking information motivation to attitudes towards Web advertising

Conference Paper Validating a Plagiarism Scale: A Middle Eastern Context

Conference Paper Exploring Mobile Advertising Effectiveness in Syria

Conference Paper Motivational Factors valued by Millennial workers in Canada

Conference Paper The effects of attitudes towards performance appraisal on or...

Article When empathy hurts: Modelling university students’ word of m...

Conference Paper E-Mail Advertising: A Middle Eastern Perspective and The Mod...

Article Teacher’s Turnover Intentions: Examining the Impact of Motiv...

Article E-mail Advertising in Syria: Assessing Beliefs, Attitudes, a...

Article Exploring Personal Experience of Wartime Crisis Effects on J...

Article A Confirmatory Factor Analysis for SERVPERF Instrument based...

Hi @Kevin, thanks for your response

My research is on the workplace environment, and this particular factor was named 'individual control over the environment'. The only two items (both scored on 7-point scales) that loaded onto it with an eigenvalue >0.4 were 'control over temperature' and 'control over lighting'. For both of those items, a high score indicated that the respondent believes individual control over the environment is very important, but for some reason that translates into a strongly negative factor score (whereas low scores on those items translates into a strongly positive score). No reverse-scored items for the factor or anything like that, which is why I found it strange - but I've repeated the factor analysis a couple of times and the exact same factor scores are generated.

Hey Kevin, my advice would be to keep in mind that the negative loadings you observe are going to be dependent on all the survey items and the way they relate to each other. Remember that your factor analysis is trying to pick apart the underlying constructs from your questionnaire. Therefore, if Factor 1 and Factor 2 are inversely related on face value, then it's likely that the loadings relating to either one of these factors will need to be negatively coded.

For instance, if you had questions that load onto a factor like "workplace stress", don't be surprised if your "control-of-temperature" questions load negatively onto their main factor, as "control over the environment" is likely inversely associated with the experience of stress.

Hope this helps, best of luck - Adam