Does labelling improve sorting?

The question at the bin

Sorting is the job a disposal label is written to do. The experimental evidence in Does packaging labelling change behaviour? points that way. This article looks at the sorting record on its own: how households sort without effective labelling, what experiments and audits show about the labels’ contribution, and what is missing.

One planning-stage claim is dropped here. A much-cited “physical bin audit showing conformance from 95% on paper to 31% on milk bottles” could not be located in any Norwegian, Danish, Dutch or British source searched, and it is not asserted. The verifiable record is population surveys, controlled experiments, and one cross-sectional audit without a baseline.

How sorting actually goes

WRAP’s Recycling Tracker is the UK population benchmark. The Spring 2025 wave surveyed 5,438 UK adults responsible for household waste, quota-sampled, in March 2025. “Over three in four (79%) of UK citizens report disposing of recyclable materials in general rubbish” — missed capture, averaging 2.5 recyclable item types per household. “81% of citizens are putting non-recyclable items in recycling bins” — contamination (WRAP, 2025). Only 9 per cent report feeling very confident about what can and cannot be recycled. Strength grade: moderate — large, well-quotaed, entirely self-reported. Self-report tends to overstate correct sorting, so the true figures are unlikely to be better.

The Australian counterpart, from Cleanaway-released 2021 research reported through Planet Ark: “just 25 per cent of Aussies are separating waste correctly at home,” with up to 35 per cent of recycling needlessly landfilled, 47 per cent wrongly believing soft plastics belong in kerbside bins, and 53 per cent thinking greasy pizza boxes recyclable (Planet Ark, 2021). Strength grade: weak — industry research with undisclosed method and sample; directionally consistent with better-documented sources.

In mature systems with established labelling schemes, most households still mis-sort in both directions. Whatever labels contribute has not eliminated pervasive error.

Labels help, conditionally

Allison et al. (2024) — 1,008 UK participants, simulated sorting of compostable formats — showed the unlabelled baseline following appearance. A “do not recycle” label lifted correct general-waste disposal by 11.15 percentage points. Worded food-waste instructions beat the bare word “compost.” The effect was significant for bags, clamshells, containers and cups, but “not for tea, sachets, sandwiches, or food stickers” — small formats where the label is physically marginal or absent. Participants judged that “a logo alone would be insufficient as a disposal instruction strategy,” which matches the design findings. Strength grade: moderate; simulated outcome.

Buelow, Lewis and Sonneveld (2010), a Melbourne mixed-methods study — 88 mail surveys and 25 interviews — found understanding of packaging labelling “often very poor,” sorting frequently incorrect despite good intentions, and action-oriented labels (“Remove cap and recycle PET”) the most effective format tested (Buelow et al., 2010). Strength grade: weak — small, 11 per cent response rate, self-report — but an early anchor for the action-oriented finding later work confirmed.

WRAP’s Boots/Radox on-pack message trial: 91 versus 86 per cent self-reported correct disposal, n = 4,000 (WRAP, 2020; 2021). Graded weak-to-moderate for self-report. The one physical-audit study located is Australian: APCO’s ARL Impact Framework bin audits — 238 households across seven New South Wales councils, October 2022, 2,356 kg of waste and 1,591 kg of recycling hand-assessed — found 62 per cent of ARL-labelled packaging in the correct stream (APCO, 2023). Strength grade: moderate for a physical audit of household bins. The report itself states the limit: it is cross-sectional, with no pre-label baseline and no unlabelled comparison, and “does not yet provide a complete picture of progress.” 62 per cent conformance describes where labelled packaging goes. It does not establish what the label contributed.

Are the labels on the packs?

Grønt Punkt Norge’s NielsenIQ audit of 200 top-selling grocery items found 63 per cent carrying the national sorting labels (86 per cent counting older or alternative symbols). Material-level variation ran from 76 per cent on plastic to 44 per cent on glass (Grønt Punkt Norge, n.d.). Strength grade: moderate — an audit of packs, not of behaviour. In a market with a long-established national scheme, roughly one top-selling package in three carried no current sorting mark. Sorting-improvement estimates that assume universal labelling inherit an availability gap before any comprehension gap begins.

Missed capture and contamination

Sorting fails in two directions at once. 79 per cent of UK households discard recyclables into general waste. 81 per cent put non-recyclables into recycling (WRAP, 2025). Missed capture is a failure of inclusion: the consumer does not know, or does not act on, recyclability. Its cost is lost tonnage, borne diffusely. Contamination is a failure of exclusion: hopeful recycling that travels downstream as sortation burden, weaker bales and rejected loads. Missed capture correlates with disengagement. Contamination correlates with engagement — “wishcycling,” the compressed everything-green-is-recyclable rule in the comprehension article.

Positive marks (“Recycle”) attack missed capture and, by lowering the threshold for inclusion, can feed contamination at the margin. Negative instructions (“Do not recycle”) attack contamination directly. The largest experimental effect in this theme is such an instruction: Allison et al.’s +11.15 points was a correct general-waste disposal gain, a contamination result, not a capture one. Conditional formats, such as the ARL’s instruction-gated tier, try to do both. A scheme’s choice among these formats is a choice about which failure to fight. Scheme documents reviewed for these articles do not state it that way.

A label change can raise capture while raising contamination, or cut contamination while suppressing capture. The Korean case in the next article is the second pattern in survey form: a quarter of respondents excluded recyclable cartons entirely. An evaluation that measures only “correct disposal” mixes two error types whose system costs differ sharply. Physical composition sampling captures both. Most survey instruments do not.

The missing evaluation

The sorting evidence is: population surveys of pervasive mis-sorting (self-report); experiments showing labels improve simulated sorting (moderate, conditional); one field trial of small self-reported improvements; one cross-sectional audit of majority-but-partial conformance. What was not found is a before/after or controlled evaluation of a label scheme’s introduction with physically measured sorting outcomes. None was located for the ARL, OPRL, Triman or the Nordic pictograms. Systematic reviews of sorting interventions, documented in What is not known?, found none either. The closest template — MRF-audited cart-tagging evaluations — has not been applied to an on-pack label change.

Schemes roll out across thousands of SKUs, so clean windows are rare. Waste audits are expensive. Schemes have more reason to publish recognition surveys than conformance experiments. An intervention increasingly mandated on sorting-improvement grounds has not been evaluated at the standard those grounds imply.

How to read scheme statistics

Outcome type, in ascending weight: stated behaviour, simulated behaviour, physical measurement. Every figure above 85 per cent in this literature is, without located exception, a stated or recognition measure. Commissioning: label-owner surveys produce the highest figures. OPRL’s 96 per cent claimed comprehension against single-digit independent definitional results is the standing illustration. Comparison structure: conformance without a baseline (the ARL audits) is one design element short of attribution. Denominator: “of those who saw the label,” “of engaged recyclers,” “of members’ packaging” convert population claims into fractions of a selected subgroup.

What the record supports

Clear, action-specific disposal instructions improve sorting decisions. That is supported by Allison, by Buelow, and by convergent design research. It is the positive claim in this theme with the largest located evidence base. That on-pack labelling improves measured household sorting outcomes is plausible and unproven. That labelling has improved system-level capture or contamination is undemonstrated. Decades of labelling coexist with 79 per cent of UK households discarding recyclables and 81 per cent contaminating. Sorting is the setting in which labelling has its largest measured effects. On the evidence assembled here those effects are positive, unquantified at population scale, and untested by physical measurement of a scheme’s introduction.

References

Allison, A.L., Băitanu, A., Purkiss, D., Lorencatto, F., Michie, S. and Miodownik, M. (2024) 'Enabling desired disposal of compostable plastic packaging: an evaluation of disposal instruction labels', Frontiers in Sustainability, 5, 1376519. Available at: Open source (Accessed: 18 August 2026).

APCO (Australian Packaging Covenant Organisation) (2023) ARL Impact Framework Report 2023, November. Available at: Open source (Accessed: 18 August 2026).

Buelow, S., Lewis, H. and Sonneveld, K. (2010) 'The role of labels in directing consumer packaging waste', Management of Environmental Quality, 21(2), pp. 198–213. Abstract available at: Open source (Accessed: 18 August 2026).

Grønt Punkt Norge (n.d.) 63 prosent av emballasjen har riktige sorteringsmerker (NielsenIQ audit), as reported at: Open source (Accessed: 18 August 2026).

Planet Ark (2021) New research on Australians' recycling behaviour shows confusion remains (Cleanaway research). Available at: Open source (Accessed: 18 August 2026).

WRAP (2020) On-pack labelling and citizen recycling behaviour. Report mirror available at: Open source (Accessed: 18 August 2026).

WRAP (2021) One simple message on pack could drive up recycling rates, press release, September. Available at: Open source (Accessed: 18 August 2026).

WRAP (2025) Recycling Tracker survey, UK, Spring 2025. Available at: Open source (Accessed: 18 August 2026).

Note on sources and verification

WRAP Tracker figures (79 per cent, 81 per cent, 2.5 items, 9 per cent confidence; n = 5,438, March 2025 fieldwork) are as stated on the survey report page. The Cleanaway/Planet Ark figures carry undisclosed methodology and are graded accordingly. Allison et al. figures are from the open-access article. Buelow et al. details (88 surveys of 800 distributed, 25 interviews) are from the publicly available abstract; the full text sits behind access controls. ARL audit figures (238 households, seven councils, tonnages, 62 per cent conformance) and the "does not yet provide a complete picture" concession are from the 2023 Impact Framework report as retrieved. The Grønt Punkt/NielsenIQ audit figures are from Norwegian trade coverage of the study; the fieldwork year was not displayed on the retrieved page. The planned "95 per cent paper to 31 per cent milk bottles" bin-audit claim was searched for in Norwegian, Danish, Dutch, Swedish and UK sources and could not be located; it is not asserted anywhere in these articles, and this note records the search so that the claim's absence is deliberate. The statement that no before/after physically measured evaluation of a label scheme exists is an absence claim, documented with the search record in this theme's final article.

Last verified: 18 August 2026.