Criminal Sentencing by Preferred Numbers

• Published date: 01 March 2020
• Author: Mandeep K. Dhami,Sheila M. Bird,Andrew McGrath,Elizabeth Merrall,Ian K. Belton
• Date: 01 March 2020
• DOI: http://doi.org/10.1111/jels.12246

Journal of Empirical Legal Studies

Volume 17, Issue 1, 139–163, Mach 2020

Criminal Sentencing by Preferred Numbers

Mandeep K. Dhami,*Ian K. Belton, Elizabeth Merrall, Andrew McGrath,

and Sheila M. Bird

Criminal sentencing is a complex cognitive activity often performed by the unaided mind

under suboptimal conditions. As such, sentencers may not behave according to policy,

guidelines, or training. We analyzed the distribution of sentences meted out in one year in

two different jurisdictions (i.e., England and Wales, and New South Wales, Australia). We

reveal that sentencers prefer certain numbers when meting out sentence lengths

(in custody and community service) and amounts (for fines/compensation). These “com-

mon doses” accounted for over 90 percent of sentences in each jurisdiction. The size of

these doses increased as sentences became more severe, and doses followed a logarithmic

pattern. Our findings are compatible with psychological research on preferred numbers

and are reminiscent of Weber’s and Fechner’s laws. The findings run contrary to argu-

ments against efforts to reduce judicial discretion, and potentially undermine the notion

of individualized justice, as well as raise questions about the (cost) effectiveness of

sentencing.

I. Introduction

Criminal sentencing represents a key stage of the criminal justice process where the

courts dispose of offenders who have pled guilty to an offense or been convicted of one.

The punishments meted out to offenders may give them their just deserts, incapacitate

or deter them from committing crimes in the future, rehabilitate them, or enable them

to make reparations. Consequently, sentencing is often predicated on the principle that

each case is unique and so dealt with on its own merits. This notion of “individualized jus-

tice” is fundamental to the sentencing framework of many jurisdictions. For instance, the

U.S. Sentencing Commission (2018:14) holds to the belief that sentences are based on

the “individualized facts of the case.” The Sentencing Council for England and Wales

*Address correspondence to Professor Mandeep K. Dhami, Department of Psychology, Middlesex University, The

Burroughs, Hendon, London, UK, NW4 4BT; email: m.dhami@mdx.ac.uk. Belton is at University of Strathclyde,

UK; Merrall is with S-cubed ApS, Denmark; McGrath is at Charles Sturt University, Australia; Bird is at University

of Cambridge, UK.

Thank you to the Sentencing Council for England and Wales, U.K. Ministry of Justice, and the New South

Wales Bureau of Crime Statistics and Research for providing access to the data sources. We also thank John Irwin

for his feedback on a draft of the article. The dataset for England and Wales (but not all of the analyses) presented

here formed part of the third author’s PhD thesis (Merrall, 2012).

139

declares that “[i]t is important to note that every criminal case is unique” (2015:9,

emphasis in original). Finally, the New South Wales, Australia sentencing bench book

(Judicial Commission of New South Wales 2014) quotes the court in R v. Engert as stating

that “[i]n every case, what is called for is the making of a discretionary decision in the

light of the circumstances of the individual case” (1995:2010).

Indeed, judicial discretion is considered essential for producing individualized jus-

tice, and jurisdictions differ in how much discretion is afforded to sentencers. In some

jurisdictions, such as in the United States, sentencers are expected to follow highly struc-

tured, numerical, grid-based guidelines that specify the factors that should determine a

sentence (e.g., offense seriousness and criminal history) and constrain the sentences that

are meted out (see Dhami et al. 2015). Other jurisdictions, such as England and Wales,

have opted for more flexible, text-based guidance that leave room for the exercise of dis-

cretion in an apparent effort to enable individualized justice (Gage 2008). Yet, in other

jurisdictions, such as in Australia, sentencers have eschewed guidelines altogether

(beyond guideline judgments) because these do not sufficiently prioritize individualiza-

tion (Australian Law Reform Commission 2006).

The exercise of discretion may, however, be problematic. Sentencers must search

for, and attend to, relevant information (e.g., harm caused, culpability of the offender),

weight it appropriately, and integrate it in order to determine a sentence. As such, sen-

tencing is a cognitive activity that may be affected by the attentional, memory, and

processing limitations of the human mind. Indeed, research has found that sentencers

may ignore or take insufficient account of legal factors (Konecˇni & Ebbesen 1982; von

Helversen & Rieskamp 2009) and may be unduly influenced by extra-legal factors

(e.g., Daly & Bordt 1995; Mitchell 2005; Steffensmeier et al. 1998). Sentencers may also

be biased by irrelevant demands or anchors (Englich et al. 2006; Guthrie et al. 2007), the

order in which cases are presented, and even the gradation of the custodial sentencing

scale used (Rachlinski et al. 2015).

In the present research, we explore the idea that sentencing decisions may be

unduly influenced by the sentencing options available to the sentencer. Sentencing

options are typically measured on continuous scales, namely, length of time (e.g., custody,

community sentence) and amount of money (i.e., fine, compensation). There is a growing

body of psychological literature on how people use continuous scales, and we consider the

implications of this for criminal punishment.

II. Preferred Numbers, Just Noticeable Differences and

Logarithmic Scaling

Psychological research has established that people use certain numbers much more fre-

quently than others when they are estimating the level of a stimulus (Baird et al. 1970;

Baird & Noma 1975; Stevens 1975). For example, when asked to generate numbers

between 1 and 100, multiples of five and 10 are more commonly selected than other

numbers (Baird & Noma 1975; Plug 1977). Multiples of five and 10 also occur more

140 Dhami et al.