Earth-retaining structures are checked twice, against two different families of failure, and the two implicate different people. External stability asks whether the wall, treated as a single coherent block, slides, tips, sinks, or rides a deeper failure surface down. Internal stability asks whether what holds the wall together holds: the reinforcement layers in a mechanically stabilized earth or segmental wall, the stem and footing in a cast-in-place one. Which family failed usually decides whose work is in question.
The external checks
Four mechanisms: sliding along the base, overturning or excessive eccentricity about the toe, bearing capacity of the founding soil, and global stability along a deep failure surface. The AASHTO LRFD bridge design specifications frame these as limit states with load and resistance factors; FHWA's MSE wall guidance and the NCMA segmental retaining wall design manual present the same family.
The inputs are geometry and soil strength: the width of the base or reinforced zone, the unit weight and friction angle of the soils, the surcharge, and the water condition. External failure is usually a question about the dimensions chosen and the parameters assumed when choosing them.
The internal checks
Internal stability asks whether load transfers through the structure. In a reinforced-soil wall, each reinforcement layer needs enough long-term tensile strength not to rupture, enough embedment beyond the assumed failure plane not to pull out, and a connection to the facing able to carry the load arriving there. In a cast-in-place wall it means the stem and footing carry the moments and shears the earth pressure generates, with the capacity ACI 318 requires.
These are per-layer checks rather than whole-wall ones. A wall can be externally generous and still fail internally at a single elevation, which is why the deformation pattern is diagnostic.
How the two look different in the field
External failure moves the wall as a unit: translation at the base with the facing still reasonably plane, rotation about the toe with the crest displacing outward, or tilt and settlement where bearing gave way. Global failure moves ground far behind and below the wall.
Internal failure deforms the wall's own face. A bulge at a particular elevation, a horizontal offset between block courses, blocks pushed out while the courses above and below stay in line, or a facing separating from the mass behind it all point inward, to reinforcement or connection.
Pullout, rupture and connection are three findings
Pullout means reinforcement was too short, too widely spaced, or embedded in soil that could not develop the assumed interaction, which is a length, spacing or backfill question. Rupture means tensile demand exceeded long-term strength, which brings in the creep, installation damage and durability reductions design guidance requires be applied to short-term index strength. Connection failure means load reached the facing and the connection to it could not carry that load.
Exhumed samples distinguish them. Reinforcement that pulled out comes back intact, abraded along its length and unbroken. Ruptured reinforcement comes back with a tensile break. Connection failure leaves the reinforcement sound and the hardware or block geometry damaged.
Why the two families implicate different parties
External stability is set by wall geometry, foundation preparation and the soil parameters used, which is the territory of the wall designer, the geotechnical engineer whose report supplied those parameters, and whoever prepared the foundation. Global stability is a site-grading question that may sit outside the wall designer's scope entirely.
Internal stability draws in the reinforcement supplier and the design-build arrangement common to proprietary systems, where the supplier furnishes the internal design and the site engineer holds only the external and global checks. Which entity held which check is a contract question worth resolving early.
Split responsibility in proprietary systems
Segmental and MSE walls are frequently delivered as vendor packages. The manufacturer's engineer designs the reinforced mass, the project engineer designs the site around it, and a specialty contractor builds it. Each interface is a place where an assumption is handed off and may never be checked.
Investigations of these walls routinely find that no single party held the whole picture. That is not a conclusion about fault, but it determines which documents matter: the vendor's submittal, the project engineer's review of it, and what conditions were transmitted between them.
Global stability is neither one
A failure surface passing beneath the wall's foundation and daylighting well beyond it is a slope-stability problem in which the wall is a passenger, and it can carry a well designed and well built wall down with it. The distinguishing evidence is the extent of ground movement, and limit-equilibrium analysis run on the full site cross-section. Analyzing only the wall section guarantees the global mechanism is missed.
What the analysis needs
Both families need as-found parameters rather than design assumptions: geometry surveyed rather than read off drawings, soils sampled and tested, reinforcement located and measured for length and spacing, and the water condition established. Back-analysis then reports a factor of safety for each mechanism against the same body of evidence.
Where several mechanisms compute close to unity, that is itself the finding. More often than not a marginal design met an adverse condition, and honest work says so rather than selecting the most attractive single cause.
Where these opinions get tested
Predictably: that the deformation pattern was read selectively, that the internal analysis used published short-term strengths without the required reduction factors, that only the wall section was analyzed, and that the responsibility split was assumed from ordinary practice rather than read from the contract documents. Work that runs both families of checks and separates the engineering finding from the contractual allocation is harder to dislodge.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.