4 MIN READ | Cognitive Psychology

Evelyn Antony

Do We Really Save the Best for Last? Here’s a Critique of O’Brien and Ellsworth’s 2012 Study

Cite This
Evelyn Antony, (2022, August 3). Do We Really Save the Best for Last? Here’s a Critique of O’Brien and Ellsworth’s 2012 Study. Psychreg on Cognitive Psychology. https://www.psychreg.org/really-save-best-last-critique-obrien-ellsworths-2012-study/
Reading Time: 4 minutes

To get the most out of anything, do we really have to save the best for last? For many, it depends on what’s going on.

In 2012, researchers Ed O’Brien and Phoebe Ellsworth, investigated whether people might judge everyday last events more positively, because they generally signal the end of an experience. Their study is one of the most cited within this area, but I’d love to explore their findings. 

When experiences come to an end, such as an individual’s favourite restaurant closing, meals appear to taste better than normal. Reminders of past experiences, such as awards ceremony, can elicit feelings of optimism. This observation led to the question of whether endings in everyday experiences, (fore example, reading the last chapter of a book), would prompt similar feelings to those retrospectively.

O’Brien and Ellsworth looked into this assertion by testing people’s ratings of different chocolates. The predictions were when the last chocolate was salient: participants would find it more enjoyable; rate it more favourably in flavour, when compared to the other chocolates; and the overall experiment would be more enjoyable. Although the predictions lead on clearly from the research question, only examples of everyday experiences were provided, rather than empirical literature, thus making it difficult to relate the present study to previous findings. 

This is an experimental study with a between-subjects and within-subjects design (for comparison of chocolate ratings). The independent variable was the saliency of the ending (theoretical), which was measured by randomly assigning participants to either the ‘last condition’ or the ‘next condition’, with the level of measurement being nominal (operational).  The dependent variable was the level of enjoyability (theoretical) measured in three ways (operational): participants’ enjoyment of the chocolates (ordinal scale 0–10), which chocolate was the best out of five choices (ordinal) and how much participants enjoyed the experiment overall (ordinal).

Although the use of ordinal scales led to easier comparisons, it was not appropriate for the present study, as neutral responses from participants (such as somewhat enjoyable) would not be magnified, while analysing the data. It was also unclear whether participants rated the overall experiment on the same ordinal 10-point scale, as the enjoyment of the chocolates or whether it was rated. 

In the double-blinded experimental procedure (participants didn’t know about flavour or number of chocolates; experimenter didn’t know about the predictions in the study), participants were given five different flavoured chocolates in a random order by the experimenter using a hidden pocket. The subjects were assigned to either the next or last condition, with the experimenter stating 1–2 phrases (‘Here’s your next chocolate,’ or ‘Here’s your last chocolate,’). The researchers collected demographic information from the participants and conducted a manipulation check to assess the study’s design, and seven incorrect responses were found during this check and data was excluded from analysis.  

The description of the procedure was easy to follow, and stringent measures were taken to reduce participation and experimenter bias. The researchers clearly identified the impact that demand characteristics – participants conforming to expectations of the research, experimenter giving cues, influencing participants’ responses – would have on the outcome of the experiment and participants’ ratings.

To improve the study, it would have been better for the researchers to explain clearly why the study was double-blinded. Although it can be assumed that the ‘next condition’ served as a control, it only assessed for differences between-subjects; and a separate control group to assess for within-subject differences in the chocolate ratings would have been appropriate. 

The results showed that there was a significant effect found between the last condition and participants’ overall ratings, the last chocolate being chosen as favourite and being rated more favourable overall. So the results follow on from the research question. A range of statistical tests were used including t-tests for between-group comparisons, a Chi-square test for comparison of nominal categories and mediation analysis using regression. Means and standard deviations of the participants were collected to compare. The statistical methods used were appropriate for the study, particularly the Chi-square test as the conditions were treated as nominal categories. Nevertheless, the authors could have presented the results more coherently and should have been clearer about the rationale for using a range of statistical tests. 

To improve the statistical methods utilised in this study, multiple comparison t-tests with an adjustment correction would be more appropriate, given the number of mean comparisons in the study. And there were seven cases of data missing from the final analysis, which could have led to reduced statistical power and consequently, the findings cannot be perceived as valid or replicable to similar studies. A pilot study could have been conducted at the start of each trial (in both conditions), so helping to identify any unexpected issues that may arise during the study. 

Overall, their study highlights the importance of time and sequential patterns, while assessing positive emotions. The results and analyses used follow on from the research question to an extent, though multiple comparison t-tests would have provided a better understanding of ratings in both conditions. The replicability of the study’s procedure comes into question, as more reasoning was required for the study being double blinded as well as the dependent variable scales not being clearly described. It would have been interesting for the researchers to focus more on the serial positioning effects and its implications on saliency of endings, in the discussion of the paper, which would in turn, provide better directions for future research in social psychology and memory. 


Evelyn Antony completed an MPhil in education at the University of Cambridge. She also finished a first class master of arts (honours) degree in psychology from the University of Edinburgh.

VIEW AUTHOR’S PROFILE


Psychreg is mainly for information purposes only; materials on this website are not intended to be a substitute for professional advice. Don’t disregard professional advice or delay in seeking  treatment because of what you have read on this website. Read our full disclaimer