A two-season maize trial in Kenya found NPK hard to beat, and mycorrhizal fungi hard to judge

In a two-season field experiment on maize in Taveta, Kenya, mineral NPK fertiliser gave the highest grain yield and the best short-term economic return of seven nutrient treatments. Adding arbuscular mycorrhizal fungi (AMF) to NPK did not significantly improve pooled grain yield over NPK alone. [3]

That is a field result, not a greenhouse result, which matters. It is also narrow. The trial ran two seasons at one site, used one maize cultivar (DK 777), and replicated each of seven treatments three times per season. Yields are reported in grams per plant, not tonnes per hectare. The abstract does not report whether the AMF inoculant successfully colonised the roots, nor what happened to soil phosphorus. Those are the measurements that would tell you whether the fungus did anything at all.

What was tested

Seven treatments: AMF alone, organic manure (OM) alone, NPK 17:17:17 alone, AMF+OM, AMF+NPK, OM+NPK, and an unamended control. Randomised complete block design, plot-level analysis, Tukey HSD for mean separation. 3

Standard agronomic design. Small scale.

The mechanism the trial was testing

AMF form symbiotic associations with plant roots and trade mineral nutrients, chiefly phosphorus, for plant carbon. A 2026 review of AMF and plant growth-promoting rhizobacteria in pistachio describes the main routes: improved phosphorus acquisition, better water relations, more stable soil structure. 1 In a semi-arid soil, those are exactly the constraints you would want relieved. Phosphorus is often locked up, and water is the limiting factor.

The theory is well supported. The Taveta trial asked something narrower and more practical: does any of it show up in maize yield, consistently, across seasons, at a price a farmer can pay?

What the trial found

Treatment had a significant effect on grain yield and biomass (P < .001). Pooled across both seasons, the highest grain yields were NPK at 148.23 g per plant, OM+NPK at 147.77 g per plant, and AMF+NPK at 146.17 g per plant. The unamended control yielded 92.03 g per plant. [3]

Those three top figures sit close enough together that the statistical test did not separate them. The authors state that pairwise combinations did not significantly outperform NPK alone for pooled grain yield. [3]

On money, NPK alone came out ahead, with the highest two-season mean net benefit (KSh 1,401.08) and the highest benefit-cost ratio (4.00). [3]

The season question is the one the paper's title hangs on. Treatment × season interactions were not significant for grain yield (P = .772) or biomass (P = .408). In plain terms, the ranking of treatments held up across the two seasons for those two variables. [3]

One below-ground trait behaved differently. Root length showed a significant treatment × season interaction (P = .011), which the authors interpret as greater seasonal sensitivity of root responses than of shoot or yield responses. [3] That is a single measured trait in a single trial, and it is a caution rather than a finding: what the roots did depended on the season, even when the yield did not.

Early vigour tracked yield. Week-6 plant height, leaf area and stem diameter were strongly associated with grain yield in both seasons. [3] That is a correlation inside this dataset, useful for anyone thinking about early-season indicators, not a causal claim.

What this does and does not support

It supports a modest, specific conclusion: at this site, in these two seasons, with this cultivar, NPK was the strongest single input, and combining it with AMF or manure did not add a statistically detectable yield advantage. The authors frame their own conclusion as support for context-specific nutrient management that weighs agronomic performance, seasonal variability and short-term returns. [3]

It does not support the broader claim that AMF do not work in semi-arid maize. The trial did not report colonisation rates, so we cannot tell whether the inoculant established. It did not report soil phosphorus, so we cannot tell whether phosphorus was limiting enough for the mycorrhizal pathway to matter. It did not report AMF-alone or OM-alone yields in the abstract, so the comparison that would isolate the fungus is not visible here. And two seasons at one site cannot establish seasonal consistency in any general sense. It can only fail to detect a difference between those two seasons.

The wider literature is consistent with that caution, and it is worth being explicit that the following sources are reviews, not new field data. A 2026 review of AMF and PGPR in pistachio notes that most studies have been conducted under controlled conditions, which limits translation of inoculation strategies to commercial orchards. [1] A critical review of plant biostimulants, a category that includes beneficial microorganisms, reports mean yield gains in the mid-teens across quantitative syntheses, with larger responses in arid, degraded or nutrient-poor soils. The same review also reports very high heterogeneity, negligible or modest responses for some commercial products in large regional field networks, poor viability of some microbial inoculants, and scarce multi-season data as constraints on confidence. 10 A review of rhizoremediation frames the same problem as a translation gap between controlled-condition biological activity and reproducible field performance. 11 A narrative review of heat and drought resilience in fruit crops argues that integrated, context-specific approaches outperform single interventions, while noting that commercial translation is constrained by microbiome instability and genotype-dependent responses. 4

None of those reviews is about maize in Taveta. They describe a general pattern, biological inputs that look strong in controlled conditions and inconsistent in fields, and the Taveta result fits that pattern.

Where the evidence runs out

The trial's own limits come first. Two seasons, one site, one cultivar, three replicates, per-plant yields. No colonisation data, no soil chemistry reported in the abstract, no AMF-alone or OM-alone figures visible there. The economic result is a two-season mean net benefit and a benefit-cost ratio. It is a short-term calculation, and the authors say so. [3]

The mechanism question is open. If the inoculant did not establish, the trial tested the absence of a fungus, not the presence of one. If soil phosphorus was already adequate, the mycorrhizal phosphorus pathway had little to offer. Neither possibility can be ruled out from what is reported.

The seasonality question is also open. Two seasons is not a test of seasonal consistency. It is a test of whether two particular seasons differed. A non-significant interaction means the data did not detect a difference, which is not the same as demonstrating stability. Root length did differ by season, which suggests the below-ground system is more responsive to conditions than the yield numbers alone imply. [3]

What would move this forward: colonisation measurements, soil phosphorus before and after, more seasons, more sites, and yield expressed per hectare so the numbers can be compared with other trials. Until then, the defensible reading is narrow. In this trial, NPK paid. Adding AMF to NPK did not measurably improve the yield or the return.