What water activity measures, and what it doesn't.

The short version: Water activity (a_w) is the ratio of a food's water vapor pressure to that of pure water at the same temperature: a measure of how much of the water is free to leave, not how much water is there [1][29]. The figure 0.85 is a scope trigger inside two specific federal regulations and a critical limit in one USDA jerky guideline, and federal documents disagree about whether it works as a general safety line at all [1][2][3][7]. A number on its own establishes nothing: every published a_w figure carries conditions, a paired pH, a temperature, a humectant, an instrument, and stripped of them it is not usable.

What water activity is

The thermodynamic definition is a ratio: "a_w = p/p_o, where a_w is the water activity derived from the laws of equilibrium thermodynamics, p is the vapor pressure of the sample, and p_o is the vapor pressure of pure water," at the same temperature and external pressure [29]. The regulatory wording in 21 CFR 113.3 is the same idea in plainer language: "the quotient of the water vapor pressure of the substance divided by the vapor pressure of pure water at the same temperature" [1]. The identical definition appears in 21 CFR 114.3 [2].

There is a second identity that matters for anyone thinking about measurement. Water activity equals the relative humidity of the air in a sealed jar around the food, divided by 100, but only "after that food has had time to equilibrate with the air" [11]. The equilibration clause is part of the definition, not a refinement of it.

Water activity also moves with temperature, and it moves in two directions at once depending on what is in the chamber. Labuza, Kaanane and Chen measured the shift on fish flour and cornmeal between 25 and 65 °C and reported that the a_w of foods increases as temperature rises, while the a_w of saturated salt solutions decreases: "This is opposite to the increase in a_w with increase in temperature for foods" [27]. A salt-calibrated instrument reading a warm sample is therefore exposed to error in two directions simultaneously.

Why moisture content is a different number

Moisture content is total water. Water activity is available water. Kansas State University Extension states the distinction directly: "Moisture content is different than a_w as it measures the total water present in a food product," while a_w "measures the amount of free water that is available in food for microorganisms to survive or grow" [12].

The strongest regulatory statement on this point is USDA FSIS on jerky. An MPR (moisture-to-protein ratio) "of 0.75:1 or less remains part of the standard of identity for jerky," but the guideline says an MPR alone is "not sufficient to ensure a safe product," because a_w "can vary significantly at any given MPR depending on solutes like salt and sugar" [7]. Two products can share a moisture figure and differ in available water, because dissolved salt and sugar bind water that a moisture measurement still counts.

Two things follow. First, the moisture targets published for home drying, "about 20 percent" for fruit and "about 10 percent" for vegetables [18], repeated as "only about 20% water" and "only about 10% water" in UGA's youth curriculum [19], are moisture targets, and neither document measures or reports water activity. Second, Purdue's freeze-drying publication records weights before and after drying and offers no method for verifying available water at home [17].

Documented ranges, with the conditions attached to each

Every row below is a published figure with the conditions its publisher attached. Where a source published a bare number with no stated conditions, the row says so: that absence is part of the record, not something to fill in.

Two entries in that table disagree inside a single document: KSU's band table places dried fruit in the 0.75–0.65 band and also places dried fruit at 15–20% moisture in the 0.65–0.60 band [12]. The moisture qualifier attaches only to the second. There is no single "dried fruit water activity" to take from it.

Freeze-dried figures are scarce, and thin where they exist. Three sources published numbers [14][16][28]; none of the three names the water activity instrument used, and one gives no measurement method at all [16]. We located no home freeze-dryer manufacturer document stating a water activity specification for finished product.

For comparison, here are the pathogen minima the FDA publishes. Every value in the first column carries one condition that travels with it: these minima are using salt as the humectant [5]. Sugar, glycerol or a polyol is a different condition, and the numbers do not transfer.

Where those two federal tables disagree, at B. cereus 0.92 against 0.93 and C. botulinum 0.935 against 0.93, we record both rather than picking one. A third federal source states C. botulinum is "unable to grow at an a_w of .93 and below" [8], which is a third formulation again. On S. aureus toxin the split is wider: one FDA document gives a flat 0.88 with no oxygen condition [6]; the FDA/IFT report, dated the same day, describes toxin production as "limited to approximately a_w 0.92 under anaerobic conditions and generally inhibited below a_w 0.90 under aerobic conditions" [3].

Moulds are where a single number is least safe. USDA ARS states most moulds "cease to grow at a water activity level below about 0.8" [8]; UC ANR gives moulds "about a_w ~ 0.65" [11]; KSU's band table places most moulds in the 0.87–0.80 band, xerophilic moulds in 0.75–0.65, and "few moulds" persisting to 0.65–0.60, with osmophilic yeasts in that lowest band [12]. USDA ARS adds that osmophilic yeasts will not grow "below 0.6" [8]. The apparent conflict is largely a "most moulds" versus "xerophilic moulds" distinction, but two of the three sources publish their number without that qualifier.

Where 0.85 comes from, and what it actually governs

There is no single "FDA 0.85 rule." There are at least three distinct federal uses of the figure, and they are different kinds of instrument.

The Food Code decision tables are pH-and-a_w grids, not a_w thresholds. Table A applies where the food was heat-treated to destroy vegetative cells and then packaged; Table B applies where it was not heat-treated, or was heat-treated but not packaged. Table B, reproduced from the FDA job aid, reads as follows, with "PA" meaning a product assessment is required [4]:

Table A, for heat-treated and packaged food, uses a coarser grid: at a_w ≤ 0.92 every pH column reads non-TCS; above 0.95 a product assessment is required at any pH above 4.6 [4]. In both tables, an a_w figure read without its paired pH is not usable.

Federal guidance also contradicts federal guidance on whether 0.85 is a safety line at all. FSIS says a limit of 0.85 "should control growth of all bacterial pathogens of concern" [7]. FDA's own contractor report says the opposite: "Technically, an a_w of 0.85 is inappropriately low as a general a_w minimum because most pathogens are inhibited at values well above 0.86" [3]. Purdue publishes two different numbers within one document: that most harmful organisms' growth is limited to a water activity of at least 0.85, and that "the minimum water activity for most unwanted organisms is 0.60" [17]. We record all of these; we do not reconcile them.

What 0.85 is not, on this record: not a shelf-stability guarantee, since no retrieved source says a_w ≤ 0.85 alone makes a food shelf stable, and both Purdue and Utah State reach for lower numbers when discussing storage [16][17]. Not a mould threshold either; xerophilic moulds and osmophilic yeasts are documented below it [8][12]. Not, by itself, a cottage food permission anywhere except where a state has written it into its own rules [10].

How it is measured, and what the instruments claim

Every accuracy figure below is the manufacturer's own specification for its own product. They were not measured under a common protocol, and three of the documents do not state their test conditions, so these figures sit side by side as claims, not as a ranking.

The spread across sensing principles inside one vendor's own line, ±0.003 for the dew-point unit against ±0.015 for the capacitance unit [20][21], is larger than the gap between several of the thresholds in the tables above. That matters when a figure is being compared to a line like 0.88 or 0.92.

The AQUALAB manual attaches three conditions that bear directly on dried product [20]. Volatiles can produce inaccurate readings, with a published diagnostic: if dew-point readings exceed capacitance readings by more than 0.018 a_w, volatiles are interfering and a different unit should be used. Sample temperature must be within ±4 °C of chamber temperature. And "extremely dry, dehydrated, highly viscous, water-in-oil emulsions (e.g., butter), high fat, or glassy" samples require extended testing or a low-emitting mode, which is a description of freeze-dried product.

Verification standards come in two kinds, and they are not interchangeable. Saturated salt solutions have published equilibrium relative humidity fixed points from the National Bureau of Standards: LiCl 11.30 ± 0.27 %ERH at 25 °C, MgCl₂ 32.78 ± 0.16, K₂CO₃ 43.16 ± 0.39, Mg(NO₃)₂ 52.89 ± 0.22, NaCl 75.29 ± 0.12, K₂SO₄ 97.56 ± 0.13 [9]. NaCl shifts slightly with temperature across the same table: 75.47 at 20 °C, 75.29 at 25 °C, 75.09 at 30 °C [9]. METER supplies a different kind: unsaturated molal solutions in single-use vials at 0.150, 0.250, 0.500, 0.760, 0.920 and 0.984 a_w, valid at 25 °C, with a stated dew-point tolerance of ±0.003 for all but the 0.150 standard [20].

What a consumer instrument can and cannot establish

The identity a_w = RH_jar/100 is real, and it is why the jar idea exists at all [11]. But the source that publishes that identity presents it conceptually and gives no hygrometer instructions, and the equilibration clause is doing load-bearing work: the reading is headspace relative humidity, which equals water activity only once the food and the air have finished equilibrating.

One extension service publishes a jar-plus-device protocol, and it is the clearest statement of the ceiling. University of Minnesota Extension names the exact device (an Elitech GSP-6, a vaccine-transport temperature and humidity logger) and surrounds it with a roughly 30-hour four-standard calibration, a two-hour equilibration hold, three to five replicate readings, an average, and a spreadsheet correction tool specific to that model [15]. It then says three things about its own protocol: "The Elitech GSP-6 was not developed for testing the water activity of food"; "Professional commercial testing is strongly recommended to verify results"; and the output is "indicative only and aimed to provide directional information when developing recipes" [15].

Virginia Cooperative Extension takes the other position: "Accurate determination of water activity requires the use of a specialized meter," no model named, with lab testing offered instead because buying such equipment "is not cost-effective for many small and mid-size producers" [13]. That document does not address consumer hygrometers at all: a silence, not a rejection. Minnesota's Department of Agriculture, separately, references "water activity self-testing kits" alongside commercial laboratories as routes a producer may take [10]. Both positions are on the record and we leave them unreconciled.

So, on what a consumer-grade reading can support:

• It is directional, by its publisher's own classification. The one extension-published protocol calls its own results indicative and recommends commercial testing to verify [15].
• It is temperature-coupled and usually uncontrolled. Food a_w rises with temperature while salt standards fall [27], and the published protocol describes no temperature control of the jar [15].
• It is device-specific. The protocol is not "read the hygrometer": the correction tool exists because the raw reading is not the answer [15].
• Its accuracy is not established here. We could not retrieve the GSP-6's manufacturer specification, and we could not retrieve the full text of the one peer-reviewed evaluation of low-cost a_w devices we located [30]. We publish no accuracy figure for any consumer device.

We are not publishing a home pass/fail protocol or a universal target number, and this is why: the thresholds in this article are paired with pH, humectant and treatment conditions, the instrument spread is wider than the gaps between several of them, and the one documented home method is classed as directional by the service that wrote it.

Documented methods, and the exact equipment each names

Each row is bound to the equipment its publisher named. None of them generalises past that equipment.

What we could not source

These are retrieval failures, not judgements that the material does not exist. Each is listed so nothing here reads as more settled than it is.

• The accuracy of low-cost devices. The full text of the one peer-reviewed evaluation we located (Cheng et al., Journal of Food Science, 2025) was not obtained across three routes [30]. The devices tested, the reference instrument, the measured bias and the stated limitations are all unknown to us. This is the largest gap in this article.
• The consumer logger's own specification. The Elitech GSP-6's stated humidity accuracy, resolution and range were not retrieved. The published correction procedure exists because of that raw specification, and we do not have it [15].
• Any freeze-dryer manufacturer water activity specification. No home freeze-dryer manufacturer document stating an a_w target or verification method was located. Claims that a home machine "achieves a_w X" have no manufacturer documentation behind them that we could find.
• Water activity values for dried culinary herbs. Nothing species-specific, basil, oregano, thyme, was obtained from a primary source. The nearest sourceable structure is spices at ~10% moisture in the 0.50 band [12], which does not separate herbs from spices. The trade method that would govern the measurement is member-restricted [25].
• An independent cross-brand instrument comparison. No non-vendor head-to-head test under a common protocol was found. Placing ±0.003, ±0.005, ±0.008 and ±0.015 side by side and calling it a ranking would be inventing one, so we have not.
• A worked equal-moisture, different-a_w example. The teaching illustration is well known; no extension, government or manufacturer source publishing the paired numeric comparison was retrieved. The distinction here is sourced qualitatively [7][12][13][17] and the underlying isotherm physics quantitatively for two commodities [27].
• The lowest a_w permitting growth of any organism. The extreme-xerophile figure often quoted could not be retrieved from a primary source, so we publish no number for it. What is sourced: xerophilic moulds in the 0.75–0.65 band, osmophilic yeasts in 0.65–0.60 [12], and osmophilic yeasts not growing below 0.6 [8].
• Two standards figures we are withholding. Novasina's SAL-T standard values were obtained without the temperature table they depend on, so we do not publish them. And the KCl values we retrieved from the NBS table are non-monotonic across 20, 25 and 30 °C, unlike NaCl over the same range [9], flagged rather than smoothed, and not republished as a standard.

Next

• When stored food fails — what the measurement is usually reached for after.
• The economics of selling what you dry — where the test becomes a cost.
• How fruit, vegetables, and meat differ — composition, which is what the reading reflects.