What Is a Hakka Roundhouse and How Does It Differ from a Fujian Tulou

A Fujian tulou is a closed circular or square rammed-earth communal fortress, shared between Hakka and Minnan groups, and it is not the same as a Hakka roundhouse. The roundhouse category is loose. It lumps together the tulou, the Guangdong weilongwu, and freestanding lineage halls in Taiwan. If you book a bus to Meizhou expecting a tulou, you will find a weilongwu and be disappointed. The difference is defensive intent, ancestral hall placement, and regional geography.

A tulou was built to be locked down. Its thick laterite-and-lime rammed-earth walls, single entrance, and inward-facing galleries create a sealed compound where hundreds of people could withstand a siege. The ancestral hall sits at the centre of the circle or square, the literal and ritual core. A weilongwu, by contrast, is open at the front. Its ancestral hall is central but the structure fans out in a semi-circle behind it, embracing a courtyard and a pond. The dragon form is a geomantic gesture, not a military one. One is a fortress that houses ancestors. The other is an ancestral house with a protective rear enclosure.

Tulou are Fujian structures. The UNESCO World Heritage inscription of 2008 covers 46 Fujian tulou sites concentrated in Nanjing county and surrounding areas, not Meizhou, not Taiwan. The technique is hangtu rammed earth using local laterite, sand, and lime, compacted in forms to compressive strengths that allow unsupported walls up to 18 metres high. The circular fortress form, with its single iron-clad gate and gun holes on upper floors, was a direct response to banditry and inter-group conflict from the late Ming through the Qing dynasty. Square tulou exist too, but the circular ones distribute structural stress better and leave no blind corners for attackers.

The communal dwelling layout puts dozens of families on the same vertical stack: ground-floor kitchens, second-floor grain storage, third and fourth storeys for living quarters. The central ancestral hall is where lineage registers are kept, rites performed, and marriages negotiated. This is not a village with a hall. It is a hall around which a village was built in a single construction campaign, often taking years. The engineering logic is inseparable from the social logic. You cannot remove the ancestor tablets and keep the structure functioning as intended.

Many tulou are presented as Ming dynasty structures, with dates like 1560 carved into lintels or recited by caretakers. The real building fabric is Qing dynasty, mostly 18th and 19th century. The gap between the Ming dynasty claim and the Qing-dynasty construction is 200 to 400 years. This is not a deliberate lie. It is lineage prestige dating: the jiapu lineage register records the founding ancestor who settled the site under the Ming, and that date gets projected onto the current structure, which is the third or fourth rebuild on the same spot. The ancestral hall may preserve a Ming-era stone threshold or a plaque, but the rammed earth around you is Qing or later.

This matters because it changes how you read the structure. A tulou dated 1560 is not a 16th-century survival. It is a 19th-century structure that replaced an 18th-century one that replaced a 17th-century one, all carrying the same lineage founding date forward. The orientation rule, the layout, and the construction technique are conservative enough that the replacement looks convincingly old. But the wet climate of southern Fujian, with its typhoons and monsoon rains, does not preserve exposed rammed earth for four centuries. The oldest standing tulou fabric is early Qing, not Ming. Accept the Qing dynasty age, and the structure makes sense. Insist on a Ming date, and you are reading a family story, not an architectural record.

Attribute	Tulou (Fujian)	Weilongwu (Guangdong)
Floor plan	Closed circle or square	Open semi-circle, dragon enclosure at rear
Primary function	Communal fortress, siege defence	Lineage dwelling, geomantic harmony
Ancestral hall placement	Exact centre of the compound	Central but forward, opening to courtyard and pond
Wall thickness at base	Up to 1.8 metres	Typically 0.6 to 0.8 metres
Regional distribution	Nanjing county, Yongding, western Fujian	Meizhou, eastern Guangdong
UNESCO status	Inscribed in 2008 as Fujian tulou	Not UNESCO-listed as a category
Ethnic association	Shared between Hakka and Minnan groups	Strongly Hakka, not shared with Minnan

The 2008 UNESCO World Heritage inscription names 46 Fujian tulou sites under the title "Fujian Tulou." The listing explicitly notes that these structures were built by both Hakka and Minnan communities, a fact often lost when the sites are marketed as purely Hakka heritage. The inscribed groups fall into three main areas: the Chuxi, Hongkeng, and Gaobei groups in Yongding county, the Tianluokeng and Hekeng groups in Nanjing county, and the Huaiyuan and Dadi groups in Hua'an county. Each contains a mix of circular fortresses and square fortresses, with the circular ones drawing the most visitors and the square ones housing more current residents.

The UNESCO citation does not use the word "Hakka" as a branding term. It specifies that the tulou are exceptional examples of communal living and defensive architecture that emerged from a specific historical context in Fujian province, not that they represent all Hakka settlement. The Minnan group, whose members built and occupied tulou alongside Hakka lineages in the same counties, receive less attention in English-language tourism material but are present in every group. If you visit and ask residents in Minnan dialects, you will find them. If you ask only in Hakka, you will miss half the story.

Two Questions, Two Building Types

A weilongwu is not a tulou with the front wall removed. It is a different typology that answers a different question. The question in Fujian was: how do we build a defensible home for hundreds of people in bandit-prone hill country where the nearest magistrate is three days away? The question in Meizhou was: how do we build a lineage house that expresses geomantic prosperity, accommodates branch families in an expandable plan, and does not need to withstand a siege? The weilongwu answers with an open courtyard plan, a semi-circular rear enclosure that mimics a dragon protecting the ancestor hall, and a front fish pond that completes the geomantic formula. It is designed to be added to as the lineage grows. A tulou is designed to be finished and filled.

Orientation Rules

The orientation rule is strict in both but operates differently. A tulou orients to the landscape defensively: the gate faces a specific direction based on topography and fengshui, but the circle itself is indifferent to direction. A weilongwu orients its ancestral hall to a precise geomantic feature, a mountain or watercourse, and the entire compound follows that axis. This is why you can find weilongwu with entrances that face any compass point, while tulou gates tend to face south or southeast for light and drainage. The orientation system in a weilongwu is legible. In a tulou it is compromised by the need for a defensible perimeter.

The Hakka roundhouse typology is not one thing. It is a family of rammed-earth communal forms that share a technique but diverge in plan, defensive character, and regional logic. The circular fortress tulou of Fujian, the square fortress tulou of the same province, the semi-circular weilongwu of Meizhou, and the freestanding gongma hall complexes of Taiwan all fall under the broader category but are not interchangeable. If you travel from Nanjing county to Meizhou expecting to see tulou, you will find weilongwu and be disappointed. If you travel the other way, you will find tulou and be confused by the absence of open courtyards.

The typology splits on two axes: defensive closure and ancestral hall centrality. A circular or square tulou scores high on both. A weilongwu scores low on defensive closure and high on hall centrality. A Taiwan gongma hall, a freestanding single-courtyard structure with the lineage shrine as the dominant element, scores low on defensive closure and moderate on hall centrality, because the hall is the structure, not part of a larger compound. The term "roundhouse" in English is itself a translation problem. It loosely translates tulou but does not capture the square tulous or the weilongwu, which is not round. Use the Chinese term for the specific form you mean, or you will book a bus to the wrong county.

Getting to Nanjing and Yongding Counties

The four main tulou groups accessible to independent visitors are in Nanjing county and Yongding county, both in Fujian province. Nanjing county is reachable by long-distance bus from Xiamen in about three hours. Yongding county takes roughly four hours from Xiamen by road. Public buses connect the county seats to the group entrances, but the last return bus leaves before 16:00. Miss it, and you are staying overnight or hiring a local driver on the spot. There is no train station at any tulou group. The nearest high-speed rail stop for Nanjing county tulou is Nanjing station, which is confusingly named: it serves the county, not the city of Nanjing in Jiangsu. Do not buy a ticket to Nanjing, Jiangsu. You will lose a full day.

Inside the Groups

Walking distances are real. The Tianluokeng group requires a steep 20-minute climb from the parking area to the viewing platform for the five-tulou panorama. The path is paved but has no shade. Go at 08:00 or after 16:00 in summer. The Chuxi group in Yongding is more spread out, with a river running through it, and you will walk two to three kilometres to cover the main structures. Entrance tickets for each group are sold at the gate. Prices for the Nanjing county sites in 2024 were tiered by group, with combination tickets available but not always cheaper than buying individually if you are only visiting two. Check the prices posted at the ticket office on the day. They change without notice.

Wrong Group, Wrong County

The most common failure is arriving at the wrong group. Tour buses from Xiamen go to the Hongkeng group in Yongding, which is the most commercialised and has the most souvenir stalls. Independent travellers who want the Tianluokeng group in Nanjing county must specify the county name to the bus driver or risk being dropped in Yongding.

Inside the Hall

The second failure is photographing the ancestral hall interior during a private rite. These halls are functioning ritual spaces, not museum exhibits. If you see incense burning, food offerings on the altar, or family members in front of the tablets, do not enter or photograph. Wait until the rite is finished or move to a different hall.

Who Lives Here

The third failure is expecting to see only Hakka people. A tulou in Nanjing county may be occupied by a Minnan lineage who have lived there for ten generations and speak no Hakka at all. Asking them in Hakka about Hakka culture will get you polite confusion. The fourth failure is treating the claimed Ming dynasty age as fact. If you repeat a 1560 date to an architectural historian, they will correct you. The fabric is Qing. Enjoy the history as lineage narrative. Do not cite it as construction date.