What Cues Do Laypeople Use to Detect Alcohol and Cannabis Intoxication?

ABSTRACT

Introduction

Alcohol and other drug use is common in Australia, yet laypeople struggle to detect intoxication in others via observation. We examined cues laypeople use to detect alcohol and cannabis intoxication to explore whether knowledge (in)accuracy may account for intoxication detection difficulties.

Methods

Participants (N = 467; 77.7% female) completed a survey assessing familiarity with/use of alcohol and cannabis. Participants reported the cues they use to detect whether a person is intoxicated with alcohol and cannabis and rated the difficulty of such detection. Researchers assessed reported cues for accuracy and organised them into descriptive categories. Participant characteristics were examined as predictors of cue accuracy.

Results

Alcohol intoxication detection was viewed as easy, while difficulty ratings for cannabis intoxication detection varied. Most reported cues to alcohol and cannabis intoxication were accurate. However, non‐responses were common across both substances. Participants provided more cues—including accurate cues—to alcohol intoxication compared to cannabis intoxication. Cannabis consumption on a ‘monthly’/‘less than monthly’ basis was associated with greater cue accuracy, and participants who viewed cannabis intoxication detection as difficult reported less accurate cannabis cues.

Discussion and Conclusions

Accurate assessment of intoxication is essential in many high‐risk contexts (e.g., healthcare, criminal justice, road and workplace safety). Our findings suggest that while laypeople can provide some accurate cues to alcohol and cannabis intoxication, substantial knowledge gaps exist—especially for cannabis. Education about an extensive range of cues to intoxication may improve intoxication detection in these contexts.

1. Introduction

Alcohol and cannabis use are widespread in Australia. In a 12‐month period, 77.0% and 11.6% of Australians aged 14 years and over consumed alcohol and cannabis respectively (Australian Institute of Health and Welfare [1]). Alcohol and cannabis consumption are associated with increases in the burden of disease (e.g., substance use disorders, overdose, chronic disease), injuries (e.g., road and workplace accidents) and violent crime (e.g., assault; [1]). Similar rates of consumption and adverse outcomes have been documented worldwide, establishing alcohol and other drug use as a pressing global issue [2, 3].

The ubiquity of alcohol and cannabis use, and their associated risks, means accurate intoxication detection is crucial in many contexts. For example, precise detection is required by health professionals to ensure a patient receives the appropriate diagnosis and treatment; by police to assess road safety and when/how an intoxicated suspect, victim or witness of a crime should be interviewed; by bartenders to determine whether a person should be served alcohol; by supervisors to ensure an employee is fit to work; and by other laypeople when deciding if their driver can safely operate a vehicle or if a sexual partner has the capacity to provide consent.

Unfortunately, objective measures of intoxication (e.g., blood alcohol concentration [BAC], urine or blood‐based drug tests) are not always available or practical in these settings [4]. For example, laypeople often do not have the authority to enforce alcohol or drug detection tests. Many cases (e.g., of sexual assault) progress to trial without BAC readings or other quantifiable intoxication evidence [5]. Individuals must therefore rely on personal knowledge about particular cues to intoxication when making assessments and associated decisions.

A substantial literature documents consistently poor intoxication detection skills among police, health professionals and laypeople [6, 7, 8, 9]. While most research focused on alcohol, similar skill deficits have been recorded for cannabis intoxication [10, 11]. However, it is not sufficient to merely establish if laypeople can detect alcohol and cannabis intoxication. We must determine how they do this to remedy poor performance. Knowledge of which cues to alcohol and cannabis intoxication are used, and whether those cues are accurate, may inform evidence‐based strategies to improve detection.

Shinar and Schechtman [11] found police generally rely on similar cues to detect intoxication across both depressant and stimulant drug classes, which is concerning considering their possible contradictory effects on the body. However, a comparison with Burns et al. [12] offers different insights into the use of distinct cues to detect alcohol and cannabis intoxication. Subjective behavioural indicators such as slurred speech, disruptive behaviour, clumsiness and mood changes were rated as the most important cues for detecting alcohol intoxication by bartenders [12]. Alternatively, police relied most on rising pulse rate—an objective measure—to detect cannabis intoxication [11]. Laypeople and police possess different capabilities and legal powers. For example, bartenders do not have the right to perform vital sign assessments or drug tests, unlike police. It is therefore important to investigate potential reliance on different cues to alcohol and cannabis intoxication within the lay population specifically, in the absence of objective assessment.

We must also consider the possible contribution of personal characteristics to cue accuracy given the subjectivity embedded in observation‐based intoxication detection. The existing literature lacks a clear response to this question. More alcohol‐related work experience (e.g., bartending) does not appear to impact the accuracy or type of cues considered important for alcohol intoxication detection [12, 13]. However, Burns et al. [12] documented a higher value placed on slurred speech as an intoxication indicator by female observers compared to males. Additionally, Martin et al. [14] observed differences in the number of standard drinks and BAC use to define low, moderate and severe intoxication depending on age and personal alcohol consumption habits. At a general level, these results suggest personal characteristics inform the way laypeople conceptualise intoxication—which may translate into cue accuracy effects.

1.1. The Current Study

Erroneous detection of alcohol and cannabis intoxication can lead to adverse, costly outcomes in many high‐risk settings. These outcomes include improper medical care; wrongful dismissal of a person's account of a crime; inaccurate judgements about whether a person should be served alcohol or is fit to work or drive; and poor assessments of whether a person has the capacity to consent to sexual activity. Many of these decisions are not made by professionals with prior training on signs of substance use. The current study therefore aimed to establish and compare cues used by laypeople to detect alcohol and cannabis intoxication, and the accuracy of these cues. We also investigated whether participants' personal characteristics (e.g., age, gender, alcohol and cannabis consumption habits, alcohol‐related training) and ratings of intoxication detection difficulty influenced the accuracy of cues reported. We did not propose directional hypotheses given the exploratory nature of the study.

2. Method

2.1. Design and Participants

This paper is a secondary analysis of data from a previous 2017 survey study of undergraduate students' perceptions of the impact of intoxication on eyewitness memory and credibility [15]. The current study focused on a subset of items probing the cues participants use to detect alcohol and cannabis intoxication and participants' demographic data. The project was approved by the University of Sydney Human Research Ethics Committee [REF: 2016/999]. Monds et al. [15] contains additional information about survey items related to eyewitness memory and credibility that are not discussed further in this paper.

The sample comprised 467 undergraduate psychology students from The University of Sydney (77.7% female, 22.3% male). All participants were above the legal drinking age in Australia (18 years). The average age was 20.5 years (SD = 5.1, range = 18–53 years). Table 1 provides a summary of participants' alcohol and cannabis consumption behaviours and alcohol‐related work/training experiences.

TABLE 1.

Distribution of substance consumption behaviours and alcohol‐related work and training experiences.

	Never, n (%)	Less than monthly, n (%)	Monthly, n (%)	Weekly, n (%)	Daily or almost daily, n (%)	Prefer not to answer, n (%)
Frequency of cannabis consumption	381 (81.6)	61 (13.1)	16 (3.4)	4 (0.9)	1 (0.2)	3 (0.6)

	M (SD)	< ‘Hazardous’ cut‐off n (%)	> ‘Hazardous’ cut‐off n (%)
Total AUDIT‐C score	3.6 (2.2)	128 (33.7)	252 (66.3)

	Yes, n (%)	No, n (%)
Alcohol‐related work experience	103 (22.1)	364 (77.9)
Alcohol‐related training	122 (26.1)	345 (73.9)

Note: Responses to Alcohol Use Disorders Identification Test—Consumption (AUDIT‐C) items were non‐compulsory and 87 participants did not provide sufficient data to generate a total AUDIT‐C score.

2.2. Measures

The following measures were extracted from the original study. Further details about additional measures beyond the scope of this study are available in Monds et al. [15].

2.2.1. Demographics

Participants provided demographic information including their age and gender.

2.2.2. Substance Familiarity

Multiple choice questions assessed familiarity with alcohol and cannabis. Personal consumption of the substance, witnessing another person consume the substance and/or general knowledge about the effects of the substance were defined as necessary to report familiarity. Participants could abstain from responding to these items.

2.2.3. Alcohol‐Related Work and Training

Participants reported whether they had worked in a role selling/serving alcohol and if they had completed training or an accreditation for this purpose. Those who responded affirmatively to the second item provided the name of the accreditation and the country it was completed in.

2.2.4. Substance Consumption Habits

The Alcohol Use Disorders Identification Test—Consumption (AUDIT‐C; [16]) is a three‐item multiple‐choice screening tool used to assess alcohol consumption habits (e.g., ‘How often do you have a drink containing alcohol?’). The following response options were available for the sample item above: ‘never’, ‘less than monthly’, ‘monthly’, ‘weekly’, ‘daily or almost daily’ and ‘prefer not to answer’. The remaining two AUDIT‐C items were also delivered on a 5‐point scale, albeit with different response options. Possible total scores ranged from 0 to 12 and higher total scores indicated greater alcohol consumption. The AUDIT‐C demonstrated acceptable reliability (Cronbach's $\alpha$ = 0.69).

Participants also responded to the following multiple‐choice item adapted from the AUDIT‐C to assess cannabis use: ‘How often do you consume cannabis?’. ¹ Only one AUDIT‐C item was adapted due to the absence of a standardised unit of cannabis. The following response options were available: ‘never’, ‘less than monthly’, ‘monthly’, ‘weekly’, ‘daily or almost daily’ and ‘prefer not to answer’. Responses were collapsed into the following categories to manage low cell counts: ‘never’, ‘less than monthly’, ‘monthly’ and ‘weekly or more’. Participants who responded ‘prefer not to answer’ were excluded from relevant analyses.

2.2.5. Difficulty of Intoxication Detection

Participants rated the difficulty of detecting alcohol intoxication on a single‐item 5‐point Likert scale (1 = extremely difficult, 5 = extremely easy): ‘How difficult do you find it to tell when someone is intoxicated by alcohol?’. This Likert scale was replicated to assess perceived difficulty of cannabis intoxication detection. Difficulty ratings were reverse scored so that higher scores indicated greater perceived difficulty.

2.2.6. Cues to Intoxication

Participants were provided with the following definition of intoxication: ‘Intoxication refers to when an individual is affected temporarily by a substance and subsequently has diminished mental and physical control’. They were asked to provide up to three visual and three verbal cues they look for when determining if someone is intoxicated by alcohol. Participants were also invited to provide a maximum of three other cues they use to detect alcohol intoxication. These three items were replicated to assess the visual, verbal and other cues participants use to determine if someone is intoxicated with cannabis. Responses were delivered via open‐text boxes. Participants could provide a maximum of nine cues to alcohol and cannabis intoxication respectively.

Each cue response received a numerical value according to the following: ‘accurate’ = 1; ‘partially accurate’ = 0.5; ‘inaccurate’ = 0. We consulted the DSM‐5 (‘Alcohol intoxication’; ‘Cannabis intoxication’; [17]), ICD‐11 (‘Alcohol intoxication’; ‘Cannabis intoxication’; [18]), and a health professional with clinical expertise in drug and alcohol services to determine cue accuracy. ² We then calculated a total cue accuracy score per substance for each participant by adding the values associated with each cue response. Total cue accuracy scores could range from 0 to 9. Cue categories are outlined below:

Accurate: Response is a recognised and observable sign of intoxication with alcohol or cannabis.

Partially Accurate: Response is highly suggestive of a recognised and observable sign of intoxication with alcohol or cannabis but does not explicitly state the sign.

Inaccurate: Response is a characteristic or feature that is not a recognised sign of intoxication with alcohol or cannabis.

The ‘visual’ and ‘verbal’ classifications were used as prompts only, particularly as participants did not always adhere to them despite providing correct cues to intoxication. Thus, we organised the reported cues to alcohol and cannabis intoxication into broader thematic categories that superseded the initial ‘visual’ versus ‘verbal’ dichotomy (see below for further explanation).

2.3. Analyses

Frequency statistics were generated for participant demographics, intoxication detection difficulty ratings, and cues to alcohol and cannabis intoxication. Cues to intoxication were coded for accuracy by a researcher (KH). A second researcher (EM) applied this coding system to 25.0% of the data and achieved satisfactory consistency (alcohol Cohen's $\kappa$ = 0.89; cannabis Cohen's $\kappa$ = 0.89).

Linear regressions were conducted to assess the impact of personal characteristics (i.e., age, gender, alcohol and cannabis consumption habits, alcohol‐related training) on the accuracy of reported cues to alcohol and cannabis intoxication (i.e., total cue accuracy scores for alcohol and cannabis). One‐way analysis of variance (ANOVA) investigated the impact of intoxication detection difficulty ratings on the accuracy of reported cues to alcohol and cannabis intoxication. Paired samples t‐tests compared the total number and accuracy of reported cues across substances. A Pearson correlation assessed whether participants who provided more accurate cues to one substance also provided more accurate cues to the other, and a Spearman correlation assessed whether those who generally reported more accurate cues across both substances also provided a greater number of cues overall.

3. Results

3.1. Perceived Difficulty of Intoxication Detection

Most participants rated alcohol intoxication detection as ‘somewhat easy’ (52.0%) or ‘extremely easy’ (29.3%). Difficulty ratings associated with cannabis intoxication detection were more varied. Most rated the task as ‘somewhat difficult’ (33.2%), followed by ‘neither easy nor difficult’ (19.9%), and ‘extremely difficult’ and ‘somewhat easy’ (both 18.6%). Figure 1 presents the distribution of difficulty ratings by substance.

FIGURE 1.

Distribution of intoxication detection difficulty ratings by substance.

3.2. Accuracy of Reported Cues to Alcohol Intoxication

Participants provided 2888 responses overall when reporting cues to alcohol intoxication. In total, 331 responses (e.g., ‘feeling tipsy’, ‘dizziness’, ‘blackout’) were excluded from further analyses because they were ambiguous, subjective (i.e., not detectable by an observer), or duplicates at the participant level. The remaining 2557 responses were coded for accuracy: 95.7% were accurate (n = 2446; e.g., ‘difficulty walking/standing’, ‘smell of alcohol on breath’), 3.8% were partially accurate (n = 96; e.g., ‘stability’, ‘smell’) and 0.6% were inaccurate (n = 15; e.g., ‘blinking often’, ‘jargon’, ‘groaning in pain’). See Figure S1 for a visual distribution of total alcohol cue accuracy scores.

Individual responses were also organised into 13 broader cue categories due to response‐level idiosyncrasies. Table 2 presents the frequency of reported cues to alcohol intoxication organised by category and accuracy. The most reported cues to alcohol intoxication described impaired speech, impaired motor control, speech (other) and behaviour (see Table 2 for sample responses).

TABLE 2.

Cues to alcohol intoxication organised by cue category and accuracy.

Cue category	Accurate (n = 2446)		Partially accurate (n = 96)
	Sample responses	n (%)	Sample responses	n (%)
Visual
Impaired motor control	Difficulty walking/standing Stumbling Loss of motor skills/coordination Loss of balance Delayed reaction time	496 (20.3)	Balance Coordination Ability to walk Stability Steadiness	26 (27.1)
Behaviour	Affection Aggression Disinhibition Hyperactivity Laughing/giggling	238 (9.7)		0 (0)
Eyes	Red/bloodshot eyes Glazed/glassy eyes Difficulty focusing eyes Drooping eyelids Dilated pupils	199 (8.1)	Eyes (whether they are focused)	1 (1.0)
Face	Red face Flushed face Exaggerated facial expressions Pale face	152 (6.2)		0 (0)
Mood	Happiness Excitedness Heightened emotions Mood swings Anger	83 (3.4)		0 (0)
Cognition	Confusion Disorientation Difficulty concentrating Inability to follow directions Memory loss	75 (3.1)	Comprehension Awareness of surroundings	5 (5.2)
Alertness	Drowsiness/tiredness Passing out Unconsciousness Falling asleep Spaced out	67 (2.7)	Alertness/eyes can track things easily	1 (1.0)
Motor control (other)	Sudden/erratic movements Animated/exaggerated gestures Limp body Slouched posture	14 (0.6)		0 (0)
Verbal
Impaired speech	Slurring Incoherent words/sentences Difficulty with pronunciation Stuttering Incomplete sentences	500 (20.4)	Cohesion Whether speech is intact	5 (5.2)
Speech (other)	Content: Admissions of drunkenness; irrational statements; offensive content Pace: Rapid or slow speech; speaking quickly or slowly Tone: Aggressive speech; argumentative speech Volume: Speaking loudly/quietly; inability to moderate volume of speech	422 (17.3)	Speed of speech Volume Fluency/speed	16 (16.7)
Communication style	Repetition of words/sentences/questions Chatting/rambling/blabbering Jumping from one topic to another	53 (2.2)	Abnormal speech Uncharacteristic speech	17 (17.7)
Other
Miscellaneous	Vomiting Alcohol/bottles in close proximity Messy/dishevelled appearance Sweating	101 (4.1)		0 (0)
Olfactory	Smell of alcohol Smell of alcohol on breath Smell of alcohol on clothes	46 (1.9)	Smell Their breath Bad breath	25 (26.0)

Note: Percentages were calculated as a proportion of accurate and partially accurate cues, respectively.

3.3. Accuracy of Reported Cues to Cannabis Intoxication

Participants provided 1835 responses overall when reporting cues to cannabis intoxication. In total, 208 responses (e.g., ‘dizzy eyes’, ‘dizzy’, ‘their composition’) were excluded because they were ambiguous, subjective or duplicates at the participant level. The remaining 1627 responses were coded for accuracy: 91.6% were accurate (n = 1490; e.g., ‘red eyes’, ‘slowed reactions’), 5.8% were partially accurate (n = 94; e.g., ‘coordination’, ‘smell’), and 2.6% were inaccurate (n = 43; e.g., ‘looking emaciated’, ‘smelling/touching surfaces’, ‘screaming’). See Figure S2 for a visual distribution of total cannabis cue accuracy scores.

Individual responses were organised into 14 broader cue categories. Table 3 presents the frequency of reported cues to cannabis intoxication organised by category and accuracy. The most reported cues described eyes, behaviour, mood and alertness (see Table 3 for sample responses).

TABLE 3.

Cues to cannabis intoxication organised by cue category and accuracy.

Cue category	Accurate (n = 1490)		Partially accurate (n = 94)
	Sample responses	n (%)	Sample responses	n (%)
Visual
Eyes	Red eyes Bloodshot eyes Dilated pupils Inflamed eyes Glazed/glassy eyes	320 (21.5)		0 (0)
Behaviour	Giggling/laughing Physical violence Hyperactivity/restlessness Aggression Irrational/erratic behaviour	155 (10.4)		0 (0)
Mood	Relaxed Happy Anxious Paranoia Chilled attitude	115 (7.7)	Change in personality	4 (4.3)
Alertness	Spaced out Tired/sleepy Lethargic Not alert Dazed	103 (6.9)		0 (0)
Motor control (other)	Slow movements Shaky Relaxed gestures Relaxed body/floppiness	73 (4.9)		0 (0)
Impaired motor control	Slowed reactions/reflexes Inhibited gross/fine motor skills Staggering/stumbling Incoordination Lack of balance	66 (4.4)	Coordination Balance	5 (5.3)
Cognition	Confusion Reduction in logic Inattentiveness Hallucinations Loss of focus/concentration	60 (4.0)	Comprehension Memory	20 (21.3)
Face	Flushed face Pale face Relaxed/blank facial expression Red face	25 (1.7)		0 (0)
Appetite	Increased appetite Eating more than usual	12 (0.8)	Hunger Munchies	30 (31.9)
Verbal
Speech (other)	Content: Offensive language Pace: Slow speech; elongated speech Tone: Happy intonation Volume: Loud speech; quiet speech	257 (17.2)	Unusual statements/comments	3 (3.2)
Impaired speech	Slurred speech Difficulty speaking Incoherent speech Can't form proper sentences Unable to talk clearly	187 (12.6)	Verbal clarity/fluency Whether speech is intact	3 (3.2)
Communication style	Trouble staying on topic Talkative/chatty Slow/delayed responses Repeating sentences/phrases	76 (5.1)		0 (0)
Other
Miscellaneous	Smoking cannabis Sweating Holding a bong/joint Evidence of marijuana paraphernalia	25 (1.7)	Dry mouth	4 (4.3)
Olfactory	Smell like cannabis Smell like weed	16 (1.1)	Smell	25 (26.6)

Note: Percentages were calculated as a proportion of accurate and partially accurate cues, respectively.

3.4. Scope of Participants' Cue Knowledge

Participants could provide a minimum of three and maximum of nine responses when describing cues to alcohol and cannabis intoxication respectively. In total, 31.3% of alcohol intoxication cue response fields (i.e., text boxes) and 56.3% of cannabis intoxication cue response fields indicated a non‐response. We included non‐responses in the following descriptive analyses due to the very high accuracy rates identified above. These analyses explored the scope of participants' knowledge of cues to intoxication: most responses provided were accurate, but what did overall response patterns tell us about participants' capacity to generate multiple accurate cues?

The proportion of accurate cues to alcohol and cannabis intoxication decreased substantially when non‐responses were included as a category in analyses. Accurate cues to alcohol intoxication dropped to 63.2%, with 2.5% of cues partially accurate, and 0.4% inaccurate. Accurate cues to cannabis intoxication fell to 37.3%, with 2.4% of cues partially accurate and 1.1% inaccurate.

3.5. Comparing Knowledge of Cues to Alcohol and Cannabis Intoxication

A paired samples t‐test compared the number of cues (of any kind) provided for alcohol and cannabis intoxication. Participants provided significantly more cues to alcohol intoxication (M = 6.2, SD = 1.5) compared to cannabis intoxication (M = 3.9, SD = 1.9), t(466) = 24.26, p < 0.001. A similar t‐test compared the accuracy of reported cues to alcohol and cannabis intoxication. Participants' total accuracy scores were significantly higher for cues to alcohol intoxication (M = 5.3, SD = 1.8) compared to cannabis intoxication (M = 3.2, SD = 1.8), t(466) = 21.53, p < 0.001.

Participants who provided more accurate cues to alcohol intoxication also reported more accurate cues to cannabis intoxication, r = 0.35, N = 467, p < 0.001. Additionally, those who generally reported more accurate cues across both substances provided a greater number of cues overall, r = 0.40, N = 467, p < 0.001.

3.6. Factors Affecting Cue Accuracy

Skewness and kurtosis statistics were observed within the desired range for alcohol and cannabis total accuracy scores (all within −0.78 and 0.12) and Q‐plots failed to document major deviations from normality.

3.6.1. Alcohol

A linear regression investigated factors affecting the total accuracy score for reported cues to alcohol intoxication. Age, gender, alcohol consumption habits (total AUDIT‐C score) and alcohol‐related training were entered as predictors and total alcohol cue accuracy score as the outcome variable. Alcohol‐related work and training responses were highly positively correlated so training was used as a proxy for both variables in these analyses, r = 0.65, N = 467, p < 0.001. Assumptions underlying the regression were assessed, with no violations detected. The regression equation was non‐significant [F(4, 375) = 1.28, p = 0.277], with an R ² of 0.01. None of the predictors were significant, all p  $\ge$ 0.083.

A one‐way ANOVA assessed the impact of difficulty ratings for alcohol intoxication detection on total alcohol cue accuracy scores. Results revealed a small significant effect of perceived detection difficulty on total cue accuracy scores, F(4, 461) = 2.40, p = 0.049, $\eta$ ² = 0.02. However, post hoc Tukey–Kramer tests adjusted for unequal group sizes did not reveal any significant comparisons between detection difficulty scale points, all p  $\ge$ 0.081. See Table 4 for a summary of the regression and ANOVA models.

TABLE 4.

Factors affecting the accuracy of reported cues to alcohol intoxication.

Variable	b	$\beta$	t	p
Age	0.01	0.02	0.41	0.684
Gender	0.31	0.08	1.42	0.156
Alcohol‐related training	−0.37	−0.10	−1.74	0.083
Personal alcohol consumption	−0.03	−0.04	−0.72	0.472

	df	F	$\eta$ 2	p
Difficulty of alcohol intoxication detection	4, 461	2.40	0.02	0.049*

Note: *indicates significance at the p < 0.05 level.

3.6.2. Cannabis

A linear regression assessed factors affecting the total accuracy score for reported cues to cannabis intoxication. Age, gender and cannabis consumption habits were entered as predictors and total cannabis cue accuracy score as the outcome variable. Cannabis consumption was dummy coded with ‘never’ as the reference group. Assumptions underlying the regression were assessed, with no violations detected except for the assumption of the normality of residuals. An observation of the P–P plot revealed non‐normality of the residuals with the datapoints forming a slight sawtooth shaped line. However, since regression is relatively robust to violations of this assumption, especially in larger samples, we proceeded with the analysis [19].

The regression equation was significant [F(5, 461) = 4.68, p < 0.001], with an R ² of 0.05. Personal consumption of cannabis on a ‘less than monthly’ basis was associated with a 0.17 unit increase in cue accuracy score compared to ‘never’. Likewise, consumption on a ‘monthly’ basis was associated with a 0.13 unit increase in accuracy score compared to ‘never’. No other predictors were significant, all p $\ge$ 0.098.

As above, a one‐way ANOVA assessed the impact of difficulty ratings for cannabis intoxication detection on total cannabis cue accuracy scores. The results indicated a medium significant effect of perceived detection difficulty on cannabis cue accuracy scores, F(4, 435) = 9.94, p < 0.001, $\eta$ ² = 0.08. Post hoc Tukey–Kramer tests adjusted for unequal group sizes revealed multiple significant differences in cannabis cue accuracy scores depending on detection difficulty ratings. Participants who rated cannabis intoxication detection as ‘extremely difficult’ provided significantly less accurate cues compared to those who rated it as ‘extremely easy’ (MD = −1.4, p = 0.013), ‘somewhat easy’ (MD = −1.5, p < 0.001), ‘neither easy nor difficult’ (MD = −1.1, p < 0.001), or ‘somewhat difficult’ (MD = −0.8, p = 0.003). Additionally, participants who rated cannabis intoxication detection as ‘somewhat difficult’ provided significantly less accurate cannabis cues compared to those who rated it as ‘somewhat easy’ (MD = −0.7, p = 0.014). No other comparisons were significant, all p $\ge$ 0.385. See Table 5 for a summary of the regression and ANOVA models.

TABLE 5.

Factors affecting the accuracy of reported cues to cannabis intoxication.

Variable	b	$\beta$	t	p
Age	0.01	0.03	0.59	0.559
Gender	0.07	0.02	0.33	0.741
Personal cannabis consumption
Less than monthly	0.95	0.17	3.78	< 0.001**
Monthly	1.29	0.13	2.80	0.005*
Weekly or more	1.35	0.08	1.66	0.098

	df	F	$\eta$ 2	p
Difficulty of cannabis intoxication detection	4, 435	9.94	0.08	< 0.001**

Note: *indicates significance at the p < 0.05 level. **indicates significance at the p < 0.001 level. The personal cannabis consumption reference group was ‘never’.

4. Discussion

Participants generally viewed the detection of alcohol intoxication as simple and cannabis intoxication as difficult. Results indicated laypeople use a variety of cues to identify alcohol and cannabis intoxication. These cues most often related to impaired speech and motor control for alcohol (e.g., slurring, loss of balance), and eyes and behaviour for cannabis (e.g., red eyes, laughing). The reported cues were predominantly accurate when non‐responses were excluded. However, the proportion of accurate cues dropped dramatically across both substances when non‐responses were included in analyses to explore the breadth of participants' knowledge. It is worth noting participants may have been aware of additional cues but chose not to report them due to factors unrelated to their cue knowledge (e.g., time constraints, limited motivation, satisfaction at providing the minimum number of responses required).

Participants reported more cues overall—including accurate cues—to alcohol intoxication compared to cannabis intoxication. Personal characteristics did not influence the accuracy of reported cues to alcohol intoxication. However, participants who consumed cannabis on a ‘monthly’ or ‘less than monthly’ basis provided more accurate cues to cannabis intoxication compared to those who had never consumed cannabis.

We also observed a relationship between intoxication detection difficulty ratings and the accuracy of cannabis cues. Participants who rated cannabis intoxication detection as ‘extremely difficult’ provided significantly less accurate cannabis cues compared to any other rating. Additionally, participants who rated cannabis intoxication detection as ‘somewhat difficult’ reported significantly less accurate cannabis cues compared to those who rated it as ‘somewhat easy’.

4.1. Accounting for Poor Intoxication Detection

Research has consistently documented poor performance when individuals are tasked with detecting alcohol and cannabis intoxication via observation [6, 7, 8, 9, 10, 13]. This literature has cited tolerance effects as one possible cause for poor detection performance: consistent substance use can increase a person's tolerance, reducing the impact of intoxication and its observable signs even at high doses [20, 21].

The current study offers an additional explanation that may also account for poor detection of intoxication in people without established tolerance, and an avenue for redress. Although participants generally reported accurate cues to intoxication, most only provided a limited number—especially for cannabis. This suggests poor detection performance may be driven by a knowledge deficit. ³ Shinar and Schechtman [11] support this interpretation: they found police typically only use one or two cues when identifying intoxication in the field.

No single cue has been established as a reliable indicator of intoxication without variation between individuals [9]. Therefore, dependence on a limited number of cues is problematic. Education is required to broaden knowledge and application of evidence‐based cues and promote a corroborative approach to detection based on multiple indicators. This education should comprise resources informed by the DSM‐5, ICD‐11, clinical knowledge, validated symptom checklists (e.g., [22, 23]), visual cue examples and opportunities for practice. Strong knowledge of cues does not necessarily imply correct application, so it is important that educational strategies incorporate practice components (e.g., sample videos).

Training should be available for free or at low cost where possible to encourage uptake beyond alcohol service contexts, which mandate certifications (e.g., Responsible Service of Alcohol) for a fee. Accessibility could also be improved by delivering training via video‐based online modules [24]. Finally, we should advocate for more frequent and practicable refresher training to facilitate knowledge maintenance.

Participants generally relied on different cues to identify alcohol and cannabis intoxication. Alcohol intoxication detection was most often associated with impaired motor control and speech, while cannabis intoxication was linked to a person's eyes and behaviour. Empirical and clinical knowledge both tell us that a broad range of cues can indicate alcohol and cannabis intoxication. We therefore need to implement substance‐specific training. Unique knowledge deficits associated with different substances must be addressed to improve comprehensive awareness of cues to intoxication and encourage corroborative approaches to detection.

The high cue accuracy rates observed in the current study, even among those without prior training, suggest knowledge of cues to alcohol and cannabis intoxication may be at least partly intuitive. However, this knowledge was consistently narrow at the individual level, indicating substantial capacity to expand lay awareness of these cues to improve poor intoxication detection.

One could also argue people outside specific occupational groups that regularly interact with intoxicated individuals (e.g., bartenders) would not benefit from cue education. However, laypeople may encounter situations in daily life where accurate detection of intoxication is crucial to their decision‐making. For example, when accepting/declining to ride in a vehicle driven by a potentially intoxicated person, or when determining if person has capacity to consent and withholding sexual advances where necessary. These scenarios can have serious consequences, underscoring the relevance of intoxication detection knowledge for all laypeople. Nonetheless, training needs may vary across populations depending on the regularity with which people encounter intoxication, intoxication‐related outcomes and their associated liability.

It is also worth noting that increasing cue knowledge alone may only improve detection for targets who are not tolerant to the effects of alcohol and/or cannabis. Observable signs of intoxication can vary between individuals with alcohol or cannabis disorders (i.e., presumed higher tolerance due to heightened consumption) compared to those who rarely consume these substances [20, 21].

4.2. Personal Characteristics and Cue Accuracy

We did not observe a relationship between personal characteristics and the accuracy of reported cues to alcohol intoxication. Alcohol consumption is common [1], so individuals may learn cues to alcohol intoxication via exposure regardless of their personal attributes. Additionally, alcohol intoxication detection difficulty ratings were noticeably skewed toward easy within our sample, perhaps masking significant effects.

Our results suggest some characteristics of the person performing a cannabis intoxication assessment may inform accuracy. Cannabis use on a ‘monthly’ or ‘less than monthly’ basis was associated with improved knowledge of accurate cues to cannabis intoxication. Individuals who use cannabis infrequently tend to do so in social settings, while heavy use is associated with solitary consumption behaviours [25, 26]. Therefore, people who use infrequently are more likely to observe others intoxicated with cannabis, which may explain their increased knowledge of discernible cues to cannabis intoxication.

Additionally, people who rated cannabis intoxication detection as difficult tended to provide less accurate cues to cannabis intoxication compared to those who believe it is easier. This finding is notable when compared to existing alcohol‐focused research, which has found either no relationship between confidence and accuracy or a tendency toward considerable overconfidence when detecting alcohol intoxication [9]. Further research is required to account for this deviation across substances. However, cannabis intoxication detection ratings were negatively correlated with cannabis consumption habits, such that higher difficulty ratings were linked to less frequent cannabis use (r = −0.25, N = 440, p < 0.001). Experience with cannabis therefore may inform the relationship between cue accuracy and beliefs about detection difficulty.

We do not advocate for altering intoxication detection education based on personal characteristics despite some significant findings. Other variables (e.g., age, gender) were not associated with cannabis cue accuracy, and we did not observe any significant relationships between personal characteristics and alcohol cue accuracy. The observed association between cannabis intoxication detection difficulty and cue accuracy should also be interpreted with caution given (i) incompatibility with prior alcohol‐based research; and (ii) the imprecision of confidence as a predictor of accuracy [9]. Instead, intoxication detection training should be comprehensive regardless of participant characteristics to ensure practical accuracy.

4.3. Limitations and Future Directions

The current study asserts laypeople possess accurate—albeit limited—knowledge of cues to alcohol and cannabis intoxication. Future research must assess this knowledge in practice and interrogate the potential discrepancy between awareness of cues and successful application in a variety of contexts (e.g., health and workplace settings, investigative interviews, hospitality). Rote recitation of accurate information does not necessarily imply a strong ability to recognise these cues in real‐world contexts. Future research should also explore participants' experiences with alcohol and cannabis beyond personal consumption or a service/sale‐related role. Additional factors including experience with addiction and/or careers in research, healthcare, pharmacy, or law enforcement may further inform knowledge of cues to intoxication.

A substantial proportion of our sample was female (77.7%). We did not observe significant gender effects with respect to the accuracy of reported cues to intoxication. However, it is possible that the gender composition of our sample affected the frequency of certain types of cues reported. For example, Burns et al. [12] observed females valued slurred speech more highly as in intoxication indicator compared to males. Future research should probe these potential differences further with a more balanced sample.

This exploratory study did not examine cues with respect to dose. Both alcohol and cannabis affect people differently depending on the quantity consumed [27, 28]. Many of the cues reported occur following different levels of substance intake, meaning some cues are more useful than others when determining intoxication in a specific case. Additionally, many contexts require laypeople to determine a degree of intoxication rather than its mere presence (e.g., bartenders deciding when someone should not be served; people determining whether it is safe to ride in a vehicle operated by someone who has consumed a substance). Monds et al. [8] documented discrepancies between observers' intoxication ratings and targets' BAC across dose, suggesting there may be a ceiling effect on observable signs of intoxication at higher levels. Future research should investigate whether our cue accuracy rates persist when laypeople provide cues with specific reference to the intoxication spectrum.

We also reiterate that we do not encourage a one‐size‐fits‐all approach to intoxication detection. Symptoms of alcohol and cannabis intoxication vary between individuals and depend on biological factors (e.g., age, gender, height and weight, tolerance) and polysubstance use [27, 28]. For example, cannabis intoxication often remains undetected when a person is also intoxicated with alcohol [11, 29]. Future research should explore if and how laypeople take these factors into account when assessing intoxication.

5. Conclusion

Accurate detection of alcohol or cannabis intoxication is essential in health, road safety, legal, hospitality, workplace and interpersonal contexts. Objective assessments are often unavailable or impractical in relevant settings, so individuals must rely on observation. The current study suggests that while laypeople may possess accurate knowledge of cues to alcohol and cannabis intoxication, this knowledge is restricted. Therefore, limited expertise may account for past evidence of poor intoxication detection skills [9, 10]. Educational strategies (e.g., low‐cost, practice‐based online modules) should address this issue by diversifying lay knowledge of accurate cues to alcohol and cannabis intoxication.

Author Contributions

Each author certifies that their contribution to this work meets the standards of the International Committee of Medical Journal Editors.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Distribution of total alcohol cue accuracy scores.

Figure S2: Distribution of total cannabis cue accuracy scores.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.