Operationalization drift in the construct of impulsivity: relevance to animal models of addiction

To the Editor:

In their recent article in Translational Psychiatry, Borruto et al. [1] capitalized on the three-criteria model of addiction originally developed for cocaine by Belin and colleagues [2, 3] and subsequently adapted for alcohol [4], to report that female rats show greater vulnerability to alcohol addiction-like behaviours than males. Adapting longitudinal approaches originally developed by these earlier studies [2, 3], Borruto et al. [1] further identified a behavioural measure, namely operant responding during post-infusion time-out periods during alcohol self-administration sessions, as a predictor of addiction-like behaviours in both sexes, which they interpreted as an index of impulsivity. In contrast, baseline levels of anxiety, novelty-induced locomotor activity, and social dominance, as measured using paradigms with established construct validity, did not predict addiction-like behaviours. Based on these findings, the authors propose impulsivity as a broad, sex-independent vulnerability factor for addiction-like behaviours, apparently consistent with previous evidence in males that impulsivity is an endophenotype of vulnerability to addiction [2, 5–7].

The investigation of sex-dependent effects in addiction is of clear importance, and the study provides valuable insights in this regard. Nonetheless, we wish to raise concerns regarding the operationalization of impulsivity in this study and, consequently, the conclusion drawn within this specific experimental context.

In this study, impulsivity was operationalized by operant responding during post-infusion time-out periods during alcohol self-administration sessions. However, this operationalization lacks sufficient construct validity. We and others have not observed a relationship between this behavioural variable and measures of impulsivity validated across species, including humans, such as premature responses in the 5-choice serial reaction time task [5-CSRTT; 2]. More fundamentally, responses emitted during time-out periods, which are generally referred to as perseverative responses, including in the 5-CSRTT, occur in the absence of programmed consequences and do not require the animal to withhold responding to obtain optimal outcomes. As such, they do not directly index failures of response inhibition, the defining feature of impulsive action. Such responses have been shown not to correlate with impulsive or addiction-like behaviors (Belin et al., 2008, 2011) [2, 8].

In addition, the interpretation of this measure as a predictor of addiction vulnerability by Borruto et al. is further limited by its temporal and experimental context. Since this measure is obtained during ongoing alcohol self-administration it is inherently influenced by chronic alcohol exposure. As such, it cannot be considered measuring an independent or premorbid trait. Predictors of vulnerability are typically defined as variables that precede, and are not themselves shaped by, drug exposure. Thus, rather than indexing impulsivity or a premorbid vulnerability marker, time-out responding may instead reflect dysregulation of alcohol-directed behaviour that contributes to the perpetuation of drug use.

These considerations are particularly relevant considering how the findings of Borruto et al. are framed. While the discussion appropriately acknowledges that time-out responding may reflect multiple overlapping processes, including persistent alcohol-seeking, reduced behavioral flexibility, and potentially impulsivity, it also notes that further work using dedicated tasks is required to disentangle these processes. In the methods, however, the same behavioural variable is explicitly used as a measure of impulsivity, in line with its reference in the abstract, which reports that impulsivity predicts addiction-like behaviours in both sexes. This creates a discrepancy between the cautious interpretation presented in the discussion and the statement made in the abstract, which, given the prominence of the abstract in shaping interpretation and citations, risks conveying that a validated measure of impulsivity was shown to be a sex-independent predictor of addiction vulnerability.

Impulsivity has gained translational relevance precisely because it is operationalized using well-established tasks specifically designed to measure inhibitory control across species. Procedures such as the 5-CSRTT, go/no-go, the fixed consecutive number schedule, the differential reinforcement of low rates of responding, and the stop-signal task, in which the inability to withhold a response leads to sub-optimal or loss of outcome, have direct analogues in tasks used in humans and therefore support cross-species translational and face validity [9]. These tasks have contributed substantially to the neurobiological framework underpinning translational research on impulsivity and have been instrumental in elucidating the neurobehavioural mechanisms linking impulsivity to increased vulnerability to addiction [10].

Nevertheless, recent studies, particularly using the three-criteria model of addiction, have increasingly interpreted time-out responding as a measure of impulsivity [1, 11–14]. While this may reflect an attempt to capture complex behavioral phenotypes within self-administration paradigms, it also risks gradually normalizing the practice, leading to a conceptual drift in the construct validity of impulsivity. This drift in how impulsivity is defined compromises the interpretability of findings and theoretical coherence in the field. This issue is particularly important considering the broader implications of the Borruto et al. [1] study. The conclusion that impulsivity predicts addiction-like behaviours in both males and females extends a well-established literature demonstrating predictive associations in males [2, 5–7], and suggests that impulsivity constitutes a generalizable, sex-independent vulnerability factor for addiction. Such a conclusion carries important theoretical weight for models of addiction vulnerability, not least because waiting impulsivity does not predict the tendency to escalate heroin self-administration in male rats in contrast with that of cocaine [15]. However, if the behavioural measure used does not possess adequate construct validity as a measure of impulsivity, then the interpretation that “impulsivity” predicts addiction severity in both sexes clearly overstates the evidence.

We therefore suggest that the findings may be more cautiously interpreted as indicating that elevated non-reinforced responding within the operant alcohol context predicts greater addiction-like behaviours. Such an interpretation remains consistent with the data while avoiding referring to a specific construct that has not been independently validated in this context. This interpretation may also warrant further examination of the relationship between individual differences in response rates at baseline (to which responding during time-outs contributes greatly) and under progressive ratio schedules of reinforcement, since the former can profoundly influence the latter, irrespective of differences in motivation.

In summary, while the study by Borruto et al. provides a valuable contribution by highlighting sex differences in addiction-like behaviours, stronger evidence based on validated and independent measures of impulsivity is required before concluding that impulsivity predicts vulnerability to developing addiction-like behaviours in both males and females. Maintaining precision in the operationalization of impulsivity is essential not only to advance our understanding of its contribution to addiction vulnerability but also to ensure interpretability, translational relevance, and theoretical coherence in the field.

Author contributions

All authors contributed to manuscript preparation. Per CRediT: Conceptualization (NG, JWD, DB); Writing – original draft (NG); Writing – review & editing (NG, JWD, DB).

Competing interests

The authors declare no competing interests.